fix(docs): correct Vertex AI migration, update Known Limitations, clarify Google providers

[junqigif]

Summary

Fixes three remaining items from issue #2289 (Google documentation issues):

1. Fix Vertex AI migration docs — The migration guides in google.md and vertex.md contradicted the code. The code (auto_client.py) says vertexai/<model> is deprecated and users should use google/<model> with vertexai=True, but the docs said the opposite.

2. Update Known Limitations — Removed outdated claims about Enum types not working and Union types being unsupported. The verify_no_unions check is a no-op stub (always returns True), and Optional types work correctly.

3. Clarify Google provider relationship — Added a clear explanation of the three provider prefixes (google/, vertexai/, gemini/) and which one is current vs deprecated.

Files changed

• docs/integrations/google.md — Provider prefix clarity, Known Limitations update, fixed Vertex AI migration
• docs/integrations/vertex.md — Updated all examples from deprecated vertexai/ prefix to google/<model> + vertexai=True, fixed migration docs, removed old SDK imports from examples

Fixes #2289

🤖 Generated with Claude Code

[jxnl]

Thanks for taking on the Google docs cleanup. I selected #2389 as the immediate merge candidate for #2289 because it fixes the same mode/limitation/syntax problems with a smaller and cleaner patch. This PR still has some potentially useful provider-prefix clarification, but it also makes broader Vertex example rewrites and adds date-stamped limitation language, so I would not merge it as-is. If you want to continue it, the best next version would be a narrower follow-up after #2389 lands, focused only on the remaining provider-prefix / Vertex migration guidance.

[jxnl]

The remaining Google provider-prefix clarification was consolidated and shipped in #2495; the other corrections were already on main via #2400. Closing as superseded.