feat(storage): add SQLite vector store, VectorStore protocol, and tests

[MbappeWU]

Motivation

• Provide a pluggable, persistent vector storage option for local-first workflows so memories can be persisted beyond process lifetime.
• Standardize the expected vector store interface so storage backends can be swapped (InMemoryVectorStore vs SQLite).
• Add basic tests and developer conveniences to validate similarity search and persistence behavior.

Description

• Introduce a VectorStore protocol and serialization helpers with dumps_vector, loads_vector, dumps_payload, and loads_payload in src/memomind/storage/.
• Add SqliteVectorStore as a SQLite-backed implementation that persists serialized vectors and payloads and uses cosine_similarity for scoring.
• Keep the existing InMemoryVectorStore and refactor MemoryStore to accept a pluggable VectorStore backend and upsert embeddings with a payload summary.
• Add documentation and project metadata (README.md, docs/SPEC.md, docs/ARCHITECTURE.md, pyproject.toml) and add .gitignore for common files.

Testing

• Ran pytest after adding tests/conftest.py to ensure the test runner can import the src path; the test suite collected 2 items and 2 passed.
• Tests include tests/test_vector_store.py which verifies InMemoryVectorStore ordering and that SqliteVectorStore persists and returns the inserted record.
• No automated test failures remain after the import path fix; all automated tests succeeded.

---

Codex Task