Lightonmuse

- support for batching in Select endpoint
- small fix in the setup script
- updates to docstrings, resources and tests

- Added `CalibratedSelect` client
- Updated dependencies

- Update client to reflect change in status codes returned upstream
- Remove workaround for upstream double encoding bug
- added support for `typical` generation mode in `Create`
- correct logprob type hint

Prepare for PyPi release

- Update setup scripts
- Add parameter to swtich reference and candidates in Select

- Add tokenize endpoint

Handle upstream update in output encoding