Use error structs instead of duplicated strings

[m1k1o]

Currently, all errors are just strings. They have all one thing in common: They identify a field that caused the error. Therefore I created DecodeError struct that hold this information.

The next batch of errors is about parsing a certain type. They all have the information about why the parsing failed and additionally what was the expected and actual value. Therefore I created ParseError.

Last similarity in all errors that I found was when handling unconvertible types. In contrast to parsing errors, they do not have any parsing error as the appropriate parser could not be even found. Therefore I created UnconvertibleTypeError for them.

This allows us to get detailed errors about what went wrong in the decoding process, it helps with the code clarity - lot of duplicate messages were joined to a single struct. And more importantly, what was my main motivation, it allows me to join the errors together when I use custom nested decoder.

[sagikazarmark]

Hey @m1k1o

Thanks a lot for this PR.

You just implemented error handling exactly how I wanted to. 😄