Python 3.13 auto dedents docstrings, breaking the parsing.

With this fix, parsing works the same as before.

Resolves #1090.

PyTorch plans to make the switch to weights_only=True for torch.load. We
already partly dealt with that in #1064 when it comes to
save_params/load_params. However, we still had a gap. Namely, when using
pickle directly, i.e. when going through __getstate__ and __setstate__,
we are still using torch.load and torch.save without handling
weights_only. This will cause trouble in the future when the default is
switched. But it's also annoying right now, because users will get the
FutureWarning about weights_only, even if they correctly pass
torch_load_kwargs (see #1090).

The reason why we use torch.save/torch.load for pickle is that those
functions are basically _extended_ pickle functions that have the
benefit of supporting the map_location argument to handle the device of
torch tensors, which we don't have for pickle. The map_location argument
is important, e.g. when saving a net that uses CUDA and loading it on a
machine without CUDA, we would otherwise run into an error.

However, with the move to weights_only=True, these torch.save/torch.load
will become _reduced_ pickle functions, as they will only support a
small subset of objects by default. Therefore, we wouldn't be able to
rely on torch.save/torch.load for pickling the whole skorch object.

In this PR, we thus move to using plain pickle for this. However, now we
run into the issue of how to handle the map_location. The solution I
ended up with is now to intercept torch's _load_from_bytes using a
custom Unpickler, and to specifically use torch.load there. That way, we
can pass the map_location and other torch_load_kwargs. The remaining
unpickling process just works as normal.

Yes, this is a private function, so we cannot be sure if it'll work
indefinitely, If there is a better suggestion, I'm open to it. However,
the function has existed for 7 years, so it's not very likely that it
will change anytime soon:

https://github.com/pytorch/pytorch/blame/0674ab7e33c3f627ca6781ce98468ec1dd4743a5/torch/storage.py#L525

A drawback of the solution is that we cannot just load old skorch nets
that were saved with torch.save using pickle.load. This is because torch
uses custom persistent_load functions. When trying to load with pickle,
we thus get:

_pickle.UnpicklingError: A load persistent id instruction was encountered, but no persistent_load function was specified.

Therefore, I had to keep torch.load as a fallback to avoid backwards
incompatibility. The bad news is that the initial problem persists,
namely that even when passing torch_load_kwargs, users get the
FutureWarning about weights_only. The good news is that users can just
re-save their net with the new skorch version and from then on they
won't see the warning again.

Note that I didn't add a specific test for this problem of loading
backwards nets from before the change, because test_pickle_load, which
uses a checked in pickle file, already covers this.

Other considered solutions:

1. Why not continue using torch.save/torch.load and just pass the
torch_load_kwargs argument to it? This is unforunately not that easy.
When switching to weights_only=True, torch will refuse to load any
custom objects, e.g. class MyModule. There is a way to prevent that,
namely via torch.serialization.add_safe_globals, but it is a ton of work
to add all required objects there, as even builtin Python types are
mostly not supported.
2. We cannot use with torch.device, as this is not honored during
unpickling.
3. During __getstate__, we could recursively go through the state, pop
all torch tensors, and replace them with, say, numpy arrays and
additional meta data like the device, then use this info to restore
those objects during __setstate__. Even though this looks like a cleaner
solution, it is much more complex and therefore, I'd argue more error
prone.
4. Don't do anything and just live with the warning: This will work -- 
until PyTorch switches the default. Therefore, we had to tackle this 
sooner or later.

Notes

While working on this, I thought that we could most likely remove the
cuda_dependent_attributes_ (which contains the net.module_,
net.optimizer_, etc.). Their purpose was to call torch.load on these
attributes specifically, but with the new Unpickler, it should also work
without this. However, I kept the attribute for now, mainly for these
reasons:

1. I didn't want to change more than necessary, as these changes are
delicate and I don't to break any existing skorch code or pickle files.
2. The attribute itself is public, so in theory, users may rely on its
existence (not sure if in practice). We would thus have to keep most of
the code related to this attribute.

But LMK if you think we should deprecate and eventually remove this
attribute.

- Add torch 2.6.0 to CI
- Remove torch 2.2.2
- Update torch install instructions, as they no longer provide conda
  packages
- Add test for new default of weights_only
- Update pickle file test artifact (explained in #1092)
- Update some comments
- Conditionally install triton 3.1 for torch < 2.6

* Add Contributing guidelines.

* Update pr number in changes.md

* Add PyTorch 2.7.0
* Dropped PyTorch 2.3.1
* Install triton 3.2 for torch 2.6
* Remove numpy<2.0 install, see #1059

As described in PEP 639, PEP 518 and PEP 621, the pyproject.toml file is a modern python tooling file which can manage multiple tasks of testing, linting, building of package, codecoverage ,etc. It has key-value configuration format.

Building the source and wheels for the library
This functionality does not have any breaking changes, I've tested it out with the python -m build command.

Linting with pylint and flake8
Linting with flake8 has no breaking changes. pylint has some changes. version 2.6 has deprecated no-space-check option with no_self_use being made optional. pylint options locally-enabled, super-on-old-class were deprecated in version 2.14. I've updated it with the relevant rules to make the lint efficient.

I've tested out both tools, they are working correctly.

Using classifiers for PyPI
When we publish packages on PyPI, classifers help categorize Python packages. These classifiers provide metadata about the package, such as its development status, intended audience, license, supported Python versions, and more. I've current added a standard set of packages for the library. Changes can be made to them as required.

Testing and Code Coverage
We can run the tests by running the pytest command with no breaking changes. Testing Code coverage has no breaking changes as well. You can check it with the pytest -v --cov=./skorch --cov-report=term-missing command.

Deprecating Manifest.in, requirements.txt and requirements-dev.txt
The functions of Manifest.in can be done by the .toml file as described in the setuptools docs here. The optional dependencies (prev. requirements-dev.txt) are listed in [project.optional-dependencies] and dependencies (prev requirements.txt) are listed in the project section.

Lastly, I have updated the gh-actions file to reflect these changes as well.

Due to #1108, pip install -r requirements.txt is no longer working.
Also, I took the liberty to bump the Python version used to 3.12.

The latest accelerate version has attributes on Accelerator that raise
an error when being accessed. This is causing an issue when trying to
Mock the Accelerator, resulting in a skorch test failing.

This PR fixes this by monkeypatching those attributes.

- drop 2.4.1
- update 2.7.0 -> 2.7.1

# Version 1.2.0

This is a smaller release, most changes concern examples and development and thus don't affect users of skorch.

## Changed

- Loading of skorch nets using pickle: When unpickling a skorch net, you may come across a PyTorch warning that goes: "FutureWarning: You are using torch.load with weights_only=False [...]"; to avoid this warning, pickle the net again and use the new pickle file (#1092)

## Added

- Add Contributing Guidelines for skorch. (#1097)
- Add an example of hyper-parameter optimization using [Optuna](https://optuna.org/) [here](https://github.com/skorch-dev/skorch/tree/master/examples/optuna) (#1098)
- Add Example for Streaming Dataset(#1105)
- Add pyproject.toml to Improve CI/CD and Tooling (#1108)

Thanks @raphaelrubrice, @omahs, and @ParagEkbote for their contributions.

**Full Changelog**: v1.1.0...v1.2.0

Release commit specific:

* Bump verison to 1.2.0
* Update CHANGES.md
* Remove workarounds that have been fixed in sklearn
Only affects tests