feat(vectordb): add Qdrant backend support

Add Qdrant as an alternative open-source vector database backend
alongside existing VikingDB/Volcengine backends.

Key changes:
- QdrantCollection: full VectorDBCollection implementation with support
  for dense/sparse hybrid search, scalar search with native order_by,
  full-text keyword search via payload indexes, multimodal search,
  aggregate data with pagination, and proper filter translation
- QdrantProject: VectorDBProject implementation for collection lifecycle
  management with configurable distance metrics and vector dimensions
- QdrantConfig: configuration model with URL, API key, gRPC, and
  timeout settings
- Backend factory: register 'qdrant' backend type in
  viking_vector_index_backend.py
- Auto-create TextIndex for text fields in create_index
- Use FilterSelector for delete_all_data instead of drop/recreate
- Paginate aggregate_data scroll to handle large collections
- Track created indexes properly in has_index

Dependencies:
- qdrant-client >= 1.9.0 (optional extra: `pip install openviking[qdrant]`)

Tests:
- Comprehensive unit tests for QdrantCollection (752 lines)
- Unit tests for QdrantProject (266 lines)

Author: mclamee

openviking/storage/vectordb/collection/__init__.py

@@ -26,3 +26,13 @@
     "LocalCollection",
     "get_or_create_local_collection",
 ]
+
+# Optional Qdrant support (requires qdrant-client)
+try:
+    from openviking.storage.vectordb.collection.qdrant_collection import (
+        QdrantCollection,
+        get_or_create_qdrant_collection,
+    )
+    __all__.extend(["QdrantCollection", "get_or_create_qdrant_collection"])
+except ImportError:
+    pass  # qdrant-client not installed