[Done] chore: apply changes for Issue #630

[github-actions]

Closes #630

Applied changes based on Issue #630.

Implemented a fix for the fallback behavior and traced the likely cause.

What was happening
generateCommitMessageOpenAI (via generateCommitMessage) had a single path through chat.completions.create, and any API error immediately returned getFallbackMessage. If that endpoint/model combo fails, it will always fallback.

Changes made

1. Added a resilient generation flow in src/api/openai.ts:

• Try openai.responses.create(...) first.
• If that fails, try openai.chat.completions.create(...).
• Only fallback if both fail.
• Relevant logic: src/api/openai.ts

2. Aligned commit-message default model with the main default model:

• defaultCommitMessageModel now equals defaultModel (gpt-5.3-codex) in src/api/openai.ts

3. Updated metadata/docs default text:

• action.yml
• docs/configuration.md

4. Updated tests to cover the new behavior (responses-first, chat fallback):

• __tests__/src/api/openai.test.ts

Validation

• npm run build passed.
• Running npm test -- __tests__/src/api/openai.test.ts in this repo fails due existing Jest ESM/top-level-await config mismatch, not from this patch.

[YiweiShen]

/codex For commit messages, please continue using a different model, as GPT-5.3-Codex is not available via the API.

[github-actions]

Updated as requested: commit-message generation now keeps a separate default model from the main Codex model.

Changed:

• src/api/openai.ts:19
  defaultCommitMessageModel is now gpt-4.1-mini (instead of mirroring defaultModel).
• action.yml:30
  Updated input description default for openai-commit-message-model.
• docs/configuration.md:24
  Updated docs default for openai-commit-message-model.

Validation note:

• Could not run build/tests in this environment because dependencies are not installed (esbuild and jest not found).