feat(cohere): add timeout and max_retries to chat generator

[Keyur-S-Patel]

Related Issues

• partially fixes #9309

Proposed Changes:

• Added timeout and max_retries to CohereChatGenerator.__init__ as keyword-only params (aligned with Anthropic pattern).
• Implemented retry behavior using provider SDK-compatible httpx transport injection:
  • sync: HTTPTransport(retries=max_retries)
  • async: AsyncHTTPTransport(retries=max_retries)
• Passed custom httpx_client into ClientV2 / AsyncClientV2 when max_retries is set.
• Kept backward compatibility for extra generation kwargs (**kwargs merged into generation_kwargs).
• Updated to_dict/from_dict expectations for timeout and max_retries.
• Updated CohereGenerator tests to reflect current inherited behavior (generation_kwargs instead of removed model_parameters).

How did you test it?

• hatch run test:unit tests/test_chat_generator.py
• hatch run test:unit tests/test_generator.py
• hatch run test:unit tests (full Cohere unit suite)

All unit tests passed on this branch.

Notes for the reviewer

• max_retries is implemented via httpx transport retries because direct max_retries constructor support on ClientV2 is not consistent across environments/versions.
• This PR also includes test alignment in tests/test_generator.py for current CohereGenerator behavior.

Checklist

• I have read the contributors guidelines and the code of conduct
• I have updated the related issue with new insights and changes
• I added unit tests and updated the docstrings
• I've used one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test:.

[Keyur-S-Patel]

@anakin87 Can you help me close this PR, its similar to the one you reviewed here

[Keyur-S-Patel]

Gentle reminder about pr review @julian-risch

[Keyur-S-Patel]

@davidsbatista can you look up this one, should be quick review

[julian-risch]

@Keyur-S-Patel Thank you for opening this pull request. The code changes look good to me! 👍 I looked into the Cohere SDK and confirmed that when using timeout and max_retries at the same time, for example CohereChatGenerator(timeout=30.0, max_retries=5), the SDK stores the timeout passed via client_kwargs and passes it to each httpx_client.request(..., timeout=30) done with the custom client.
Thank you also for your patience with regard to this review. I'd like to assure you that we do our best to review every community PR in a timely manner. We auto-rotate reviews and there is no need to tag individual team members in PR comments. Thank you!

[Keyur-S-Patel]

Tysm for reviewing @julian-risch