feat(compactor HS): use streaming RPCs for interacting with job queue over grpc

[sandeepsukhani]

What this PR does / why we need it:
The PR actually contains multiple changes:

1. Move the proto definition for job queue RPCs to the existing compactor client protos. This would allow us to reuse the existing compactor client and its configs.
2. I changed RPCs to streaming RPCs for interacting with the job queue over grpc to avoid polling for new jobs. I took some inspiration from our existing code for query processing using query-frontend/scheduler and queries.
3. I have made the job builder loop not return an error when it sees a failure in building the job. It will continue to attempt job building with the next tick event. We will add a metric to track the failures and alert when multiple consecutive failed attempts occur.

Checklist

• Tests updated