perf: skip redundant index creation in Neo4jDriver when indices already exist

[carlos-alm]

Summary

• Add a fast path to Neo4jDriver.build_indices_and_constraints(): query SHOW INDEXES YIELD name (1 roundtrip) and skip creation if all 31 expected indices already exist
• Add an instance-level _indices_verified flag so repeated calls within the same process are free
• Extract get_neo4j_expected_index_names() from existing CREATE queries (regex over single source of truth — no separate manifest)
• Graceful fallback: if SHOW INDEXES fails, falls through to the original 31-query behavior

Scenario	Before	After
All indices exist	31 queries (~1-2s)	1 query (~50ms)
Some missing	31 queries	32 queries (1 check + 31 creates)
Second call, same instance	31 queries	0 queries (flag)
delete_existing=True	unchanged	unchanged
SHOW INDEXES fails	N/A	falls through to 31 queries

Files changed

File	Change
graphiti_core/graph_queries.py	Add get_neo4j_expected_index_names() + regex
graphiti_core/driver/neo4j_driver.py	Fast-path logic + _indices_verified flag
tests/driver/test_neo4j_driver.py	New test file (8 tests)

Test plan

• get_neo4j_expected_index_names() returns exactly 31 index names
• Fast path: all indices exist → 1 query only
• Partial missing → falls through to creation
• delete_existing=True → bypasses fast path
• SHOW INDEXES failure → graceful fallback to individual queries
• _indices_verified flag → second call is free (0 queries)
• delete_existing=True → resets verified flag
• Ruff format/lint clean, Pyright 0 errors

[danielchalef]

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

[carlos-alm]

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