feat: Vertex AI credentials support and LangChain migration

[lfnovo]

Summary

• Add credentials_file parameter to Vertex AI provider with full credentials priority chain (explicit file → GOOGLE_APPLICATION_CREDENTIALS → ADC → gcloud CLI fallback)
• Migrate to_langchain() from deprecated ChatVertexAI (langchain-google-vertexai) to ChatGoogleGenerativeAI (langchain-google-genai)
• Remove langchain-google-vertexai dev dependency
• Bump version to 2.19.0

Test plan

• All 44 vertex provider tests pass (including 12 new tests)
• Full test suite passes with no regressions (992 passing)
• No remaining references to langchain-google-vertexai in source code
• uv sync --all-extras succeeds with updated lock file

[cubic-dev-ai]

1 issue found across 7 files

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="src/esperanto/providers/llm/vertex.py">

<violation number="1" location="src/esperanto/providers/llm/vertex.py:103">
P1: When `credentials_file` is explicitly provided, loading failures (missing `google-auth` or invalid file) should raise an error instead of silently falling back to other methods. This prevents using incorrect identities (e.g., personal gcloud credentials) when the intended service account fails to load.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}