feat: add VertexRagRetrievalTool for Vertex AI RAG Engine grounding

[nuthalapativarun]

Description

Closes #152

Adds VertexRagRetrievalTool, a new built-in tool that grounds model responses using a Vertex AI RAG corpus. Mirrors the pattern established by GoogleSearchTool — it modifies the LLM request config server-side via processLlmRequest and requires no local execution.

Key design decisions

• One tool = one corpus: The Vertex AI RAG Engine backend only supports one corpus per ragResources array. Users who need multiple corpora should create multiple tool instances.
• Follows GoogleSearchTool pattern: Extends BaseTool, overrides processLlmRequest, runAsync is a no-op.
• Optional config: similarityTopK and vectorDistanceThreshold are optional; only set when provided.

Changes

• core/src/tools/vertex_rag_retrieval_tool.ts — new VertexRagRetrievalTool class + VertexRagRetrievalToolConfig interface
• core/src/common.ts — export VertexRagRetrievalTool and VertexRagRetrievalToolConfig
• core/test/tools/vertex_rag_retrieval_tool_test.ts — 7 unit tests covering all config paths

Testing

npx vitest run core/test/tools/vertex_rag_retrieval_tool_test.ts
✓ 7 tests passed

Checklist

• I've read the CONTRIBUTING.md guide
• My code follows the code style of this project
• I've added tests that prove my fix or feature works
• All new and existing tests pass
• I've updated documentation (TSDoc) as needed