Fix/falkordb group id race condition

[kraft87]

Summary

Fix a race condition where FalkorDB driver state was mutated in-place when group_id differed from the current database, causing concurrent calls to corrupt each other's database context.

Type of Change

• Bug fix
• New feature
• Performance improvement
• Documentation/Tests

Objective

When add_episode or add_episode_bulk was called with a group_id different from the FalkorDB driver's current database, the code mutated shared state:

self.driver = self.driver.clone(database=group_id)
self.clients.driver = self.driver

Under concurrent calls (e.g., asyncio.gather with different tenants), coroutines would overwrite each other's driver, writing data to the wrong database.

Changes

Core fix -- driver isolation (graphiti_core/graphiti.py):

• Removed the in-place self.driver / self.clients.driver mutation from both add_episode and add_episode_bulk
• Both methods now accept an optional driver kwarg, create a local clients = self.clients.model_copy(update={'driver': driver}), and thread that through all internal calls
• All internal helpers (_process_episode_data, _extract_and_dedupe_nodes_and_edges, _resolve_nodes_and_edges_bulk, _get_or_create_saga, _extract_and_resolve_edges) updated to accept and propagate driver
• Every self.driver / self.clients reference in these methods replaced with the local driver / clients

Decorator-based group_id handling (graphiti_core/decorators.py):

• Added @handle_single_group_id -- detects FalkorDB provider, clones the driver scoped to the group_id, calls ensure_database_initialized(), and injects as driver kwarg. The original self.clients.driver is never touched.
• Updated @handle_multiple_group_ids -- now calls ensure_database_initialized() on each cloned driver

FalkorDB driver (graphiti_core/driver/falkordb_driver.py):

• clone() now uses copy.copy(self) instead of constructing a new FalkorDriver, avoiding re-running __init__ (which schedules index builds, creates new operation objects, etc.)
• Added _initialized_databases: set[str] -- shared by reference across all clones via shallow copy
• Added ensure_database_initialized() -- idempotent guard that calls build_indices_and_constraints() only once per database across all clones

Base driver (graphiti_core/driver/driver.py):

• Added no-op ensure_database_initialized() to GraphDriver base class

Build/test infra:

• Makefile: Added test-integration target, added DISABLE_NEO4J=1 and -n auto to unit test command, increased Docker wait timeout to 300s for WSL
• pytest.ini: Added norecursedirs = mcp_server server to prevent collecting tests from sub-projects
• Deleted empty mcp_server/tests/__init__.py

Formatting (auto-applied by make format):

• Minor reformatting in edges.py, nodes.py, community_operations.py, edge_operations.py, gliner2_client.py, gliner2_neo4j.py (line length, quote style)

Testing

• Unit tests added/updated
• Integration tests added/updated
• All existing tests pass

File	Tests	Type
tests/test_decorators.py	17	Unit -- covers get_parameter_position, @handle_multiple_group_ids, @handle_single_group_id
tests/test_group_id_isolation.py	9 unit + 2 integration	Verifies original driver is never mutated, cloned drivers are injected, concurrent calls are isolated, add_episode_bulk propagates driver
tests/driver/test_falkordb_driver.py	7 unit + 1 integration	clone(), ensure_database_initialized(), shared _initialized_databases

Breaking Changes

• This PR contains breaking changes

Checklist

• Code follows project style guidelines (make lint passes)
• Self-review completed
• Documentation updated where necessary
• No secrets or sensitive information committed

[danielchalef]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[kraft87]

I have read the CLA Document and I hereby sign the CLA