Add OpenAI-compatible embedding client to vmcp optimizer

[gabrielcosi]

Summary

The vMCP optimizer's semantic tool discovery only supported the HuggingFace Text Embeddings Inference (TEI) API, so it could not use OpenAI-compatible embedding services such as OpenAI, Azure OpenAI, or gateways like Bifrost and LiteLLM.

• Add an openai embedding provider alongside the existing TEI backend in the optimizer's similarity package.
• Add two optional optimizer config fields: embeddingProvider (enum tei/openai, default tei) and embeddingModel.
• Read the API key from the OPENAI_API_KEY environment variable so it never lands in a CRD spec or ConfigMap; an empty key allows keyless in-cluster gateways.
• Validate the provider and resolve the key in the single existing GetAndValidateConfig path; regenerate CRD manifests/docs and add an example.

Closes #5305

Type of change

• New feature

Test plan

• Unit tests (task test)
• Linting (task lint-fix)
• Manual testing (describe below)

Validated end-to-end against a live cluster: configured the openai provider to point at a Bifrost OpenAI-compatible gateway that proxies a TEI backend, and confirmed a 2-input batch returns the OpenAI response shape (ordered data[].index, float data[].embedding arrays, 384-dim), which openAIClient.embedChunk decodes correctly. Also exercised keyless and Bearer-auth request paths via unit tests.

API Compatibility

• This PR does not break the v1beta1 API (adds two optional, defaulted fields to OptimizerConfig).

Does this introduce a user-facing change?

Yes. The vMCP optimizer can now point semantic tool discovery at an OpenAI-compatible /embeddings endpoint via optimizer.embeddingProvider: openai (plus embeddingModel, and OPENAI_API_KEY). The provider defaults to tei, so existing configurations are unchanged.

Special notes for reviewers

• The API key is intentionally not a config field; it is read from OPENAI_API_KEY so the secret stays out of the CRD spec / ConfigMap (mirrors the THV_SESSION_REDIS_PASSWORD pattern).
• The openai provider reads embeddingService directly and is not used with embeddingServerRef (which provisions a managed TEI server).
• Scope is intentionally self-contained to the optimizer + config; no operator controller or CLI changes.

Large PR Justification

• Auto-generated code (CRDs) and tests.

---

Assisted by Claude Code. Design, code, and tests were directed, reviewed, and manually verified by the author.

[codecov]

Codecov Report

❌ Patch coverage is 88.46154% with 12 lines in your changes missing coverage. Please review.
✅ Project coverage is 70.37%. Comparing base (e7934ea) to head (dd1ddf5).
⚠️ Report is 9 commits behind head on main.

Files with missing lines	Patch %	Lines
...mcp/optimizer/internal/similarity/openai_client.go	83.33%	6 Missing and 6 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5633      +/-   ##
==========================================
+ Coverage   70.33%   70.37%   +0.03%     
==========================================
  Files         648      651       +3     
  Lines       66011    66210     +199     
==========================================
+ Hits        46432    46598     +166     
- Misses      16221    16258      +37     
+ Partials     3358     3354       -4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jerm-dro]

Thanks for the contribution, @gabrielcosi — clean implementation and thorough tests. A few minor changes requested inline.

[github-actions]

Large PR Detected

This PR exceeds 1000 lines of changes and requires justification before it can be reviewed.

How to unblock this PR:

Add a section to your PR description with the following format:

## Large PR Justification

[Explain why this PR must be large, such as:]
- Generated code that cannot be split
- Large refactoring that must be atomic
- Multiple related changes that would break if separated
- Migration or data transformation

Alternative:

Consider splitting this PR into smaller, focused changes (< 1000 lines each) for easier review and reduced risk.

See our Contributing Guidelines for more details.

---

This review will be automatically dismissed once you add the justification section.

[gabrielcosi]

@jerm-dro the large-PR bot is giving me side-eye, although most of the lines are generated CRDs/docs and tests. Want me to add the justification?

[jerm-dro]

@gabrielcosi Adding the justification to the PR description is fine! It's common to say something along the lines of "most of these changes are auto-generated or tests."

Large PR justification has been provided. Thank you!

[github-actions]

✅ Large PR justification has been provided. The size review has been dismissed and this PR can now proceed with normal review.

[jerm-dro]

Thanks for the contribution! 🚀