[FEATURE] Add ChromaDB as builtin target (Phase 1 MVP)

[ambicuity]

Summary

This PR implements ChromaDB as a built-in target for CocoIndex, following the pattern established by the LanceDB target. This addresses issue #1214.

Changes

Core Implementation

• Added python/cocoindex/targets/chromadb.py with full ChromaDB connector
• Implements core CRUD operations (create, read, update, delete)
• Schema mapping from CocoIndex types to ChromaDB documents
• Support for vector embeddings storage
• Proper error handling with helpful messages for missing dependencies

Dependencies

• Added chromadb>=0.5.0 as optional dependency in pyproject.toml
• Uses guarded import pattern with clear error messaging

Features (Phase 1 MVP)

• Basic ChromaDB connector with persistent and in-memory client support
• Single key field support (matching LanceDB pattern)
• Conversion of Python types to ChromaDB-compatible formats
• UUID, Range, Vector, Struct, and Table type support
• Metadata storage for non-vector fields
• Embedding storage for vector fields
• Collection lifecycle management (create/delete/reuse)

Testing

Manual testing performed with ephemeral ChromaDB client. Comprehensive automated tests will be added in a follow-up PR.

Future Work (Phase 2)

As discussed in #1214:

• Vector index configuration with distance metrics
• Advanced indexing options
• Performance optimizations
• Query capabilities

Checklist

• Implementation follows LanceDB target pattern
• Optional dependency properly configured
• Error messages are helpful and actionable
• Code includes docstrings
• Single key field requirement enforced
• Tests added (deferred to follow-up)
• Documentation updated (deferred to follow-up)

Fixes #1214