fix: use correct storage path in run_benchmark.py

[Aneesh-382005]

Summary

benchmarks/run_benchmark.py was missed in #106. It still uses Path(config.storage_path) / project_dir.name instead of project_storage_dir(), so it looks in a different directory than where the index is written (<name>-<hash>). Because of that, every benchmark query returns 0 results.

• Found this while producing new results for Add benchmark for Django repo #27 and Add benchmark for Express.js repo #28 every query
  came back as a MISS with 0 tokens until I traced it to this mismatch.

Fix

Use the same project_storage_dir() helper the indexer already uses.

Test plan

• Ran python benchmarks/run_benchmark.py --repo https://github.com/expressjs/express.git --source-dir lib before the fix: all 20 queries MISS, 0% everywhere
• Same command after the fix: 20/20 HIT, 93.6% retrieval savings, 97.1% combined
• Note for maintainers: benchmarks/results/fastapi.md, fiber.md, chi.md were generated before fix: hashed storage paths to prevent basename collisions #106 landed and are stale relative to current main, though still historically valid. Happy to rerun them if useful.

[rajkumarsakthivel]

Closing as superseded by #136, which includes this fix along with the Django and Express benchmark results. The storage path fix is now on main.