feat(storage): add Milvus vector adapter

[zc277584121]

Summary

• Add a Milvus vector database adapter with Milvus Lite defaults and configurable URI, token, database, and consistency level.
• Register the backend in the VectorDB adapter factory and add configuration/schema support.
• Document Milvus Lite, self-hosted Milvus, and Zilliz Cloud configuration examples.

Testing

• uvx ruff format --check openviking/storage/vectordb_adapters/milvus_adapter.py tests/storage/test_milvus_adapter.py openviking_cli/utils/config/vectordb_config.py
• uvx ruff check openviking/storage/vectordb_adapters/milvus_adapter.py tests/storage/test_milvus_adapter.py openviking/storage/vectordb_adapters/factory.py openviking/storage/vectordb_adapters/__init__.py openviking_cli/utils/config/vectordb_config.py
• python3.10 -m py_compile openviking/storage/vectordb_adapters/milvus_adapter.py tests/storage/test_milvus_adapter.py openviking_cli/utils/config/vectordb_config.py
• git diff --check origin/main...HEAD
• uv lock --check
• PYTHONPATH=. uv run --no-project --python 3.10 --with pytest --with pydantic --with typing-extensions --with pyyaml --with httpx --with loguru --with 'pymilvus[milvus_lite]' --with apscheduler --with json-repair --with xxhash --with volcengine pytest --confcutdir=tests/storage -o addopts= tests/storage/test_milvus_adapter.py -q

[zc277584121]

Hi @qin-ctx @ZaynJarvis, could you take a look when you have a chance?

This PR adds a Milvus vector database adapter using the existing CollectionAdapter/factory structure. The branch has been rebased on the latest main and is mergeable again. Focused local checks are passing: py_compile, ruff check, ruff format --check, uv lock --check, and git diff --check.

[zc277584121]

Rebased this branch onto the latest main and resolved the vector-backend conflicts against the current adapter architecture. The retired Qdrant/OpenGauss configuration was not reintroduced.

Additional fixes from server validation:

• preserve Milvus COSINE similarity scores (higher is better) instead of treating them as distances
• use a server-compatible 2D metadata vector
• create and load an AUTOINDEX for the metadata collection, including repair of an existing unindexed metadata collection
• keep the updated dependency lock and configuration documentation aligned with the current tree

Validation completed:

• 11 focused Milvus tests passed, including the Milvus Lite create/upsert/filter/search/delete smoke test
• Zilliz Cloud passed metadata persistence, create, reconnect, upsert, filtered search, count, and delete
• Cloud smoke collections were verified cleaned up
• ruff check and ruff format --check passed for all changed Python files
• uv lock --check and git diff --check passed

A fresh local source build of the latest main currently segfaults while initializing the native vectordb engine (_new_schema) before any Milvus adapter code is reached. The focused Milvus suite was therefore also run with that unrelated native backend isolated; both the Lite and Cloud paths completed successfully.