fix: resolve SQLite embedding ValueError for list type (fixes #382)

[octo-patch]

Fixes #382

Problem

When using the SQLite backend, table creation fails with:

ValueError: <class 'list'> has no matching SQLAlchemy type

The SQLite model classes (SQLiteMemoryItemModel, SQLiteResourceModel, SQLiteMemoryCategoryModel) inherit embedding: list[float] | None from parent Pydantic models. When build_sqlite_table_model creates table models with table=True, SQLModel attempts to map all fields to SQL columns. Since SQLite has no native vector type and list has no default SQLAlchemy mapping, this raises a ValueError.

The PostgreSQL models avoid this by re-declaring embedding with sa_column=Column(Vector(), ...), but the SQLite models used a separate embedding_json field + @property pattern that left the parent embedding field unmapped.

Fix

• Add a JSONEncodedList SQLAlchemy TypeDecorator that serializes list[float] to JSON text on write and deserializes back on read
• Override the parent embedding field in each SQLite model with sa_column=Column(JSONEncodedList(), nullable=True), matching the PostgreSQL approach
• Remove the now-unnecessary embedding_json field and @property getter/setter boilerplate
• Update repository code to use row.embedding directly (the TypeDecorator handles serialization transparently)

Test

The fix can be verified by running python tests/test_sqlite.py — table creation should no longer raise ValueError.