fix: include max_tokens for OpenAI-compatible chat

[devlikebear]

Summary

• Fixes OpenAI-compatible chat clients ignoring ProviderOptions.MaxTokens.
• This belongs in TARS because downstream apps such as Linetta should configure provider behavior through TARS instead of shipping provider-specific chat clients.

Changes

• Main user-visible or developer-visible changes:
  • OpenAI-compatible chat requests now include max_tokens when ClientConfig.MaxTokens is configured.
  • Added regression coverage for the serialized /chat/completions payload.
• API, config, or compatibility changes:
  • No API shape changes; existing ProviderOptions.MaxTokens now applies to OpenAI-compatible providers.

Validation

• make lint-diff
• make test-cover-diff
• make test
• make security-scan
• Additional manual verification, if needed

Checklist

• Commit messages follow feat/fix/chore
• Tests added or updated for behavior changes
• Docs and CHANGELOG.md updated when needed
• If this is a release PR, VERSION.txt and CHANGELOG.md changed together
• Breaking changes include migration notes
• No secrets, private keys, or local absolute paths included

Risks / Rollback

• Risk level: Low. This only emits an optional request field when callers explicitly configured MaxTokens.
• Rollback plan: Revert this commit to omit max_tokens again for OpenAI-compatible chat requests.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud