perf(falkordb): add auto-index creation on match_keys in upsert_docs_batch

[JulienDbrt]

Context

Following the Memgraph connector improvements in PR #131, we identified that upsert_docs_batch performance degrades significantly without indexes on match_keys.

The MERGE operation without an index on the match keys results in O(n) scans per document, leading to O(n²) overall complexity for batch inserts.

Proposed Solution

Add automatic index creation on match_keys in upsert_docs_batch for FalkorDB, similar to what was implemented for Memgraph:

# Auto-create index on match_keys for MERGE performance (idempotent)
for key in match_keys:
    try:
        self.execute(f"CREATE INDEX FOR (n:{class_name}) ON (n.{key})")
    except Exception:
        pass  # Index already exists

Expected Impact

Based on Memgraph results:

• Before: ~1000 docs/sec for large batches
• After: ~72,000 docs/sec (70x improvement)

Notes

• FalkorDB uses slightly different Cypher syntax: CREATE INDEX FOR (n:Label) ON (n.property)
• Index creation is idempotent (errors on existing index are silently ignored)

[alexander-belikov]

That's a good catch and makes total sense.

I tend to think that explicit is better than implicit in this case : the necessary indexes should be declared in the schema.

Perhaps a higher abstraction should be sent to upsert_docs_batch and generate a warning in case we are trying to ingest docs on keys that are not indexes.