Bench/django express queries

[Aneesh-382005]

Summary

Closes #27 and #28
Adds benchmark results for Django (Python, large/deep-hierarchy repo)
and Express (JavaScript, tests tree-sitter JS chunking), per #27 and #28.

Depends on #134 : that PR fixes a storage-path bug in
run_benchmark.py that makes every query return 0 results. This PR's
numbers were produced with that fix applied locally; please merge #134
first (or review this diff assuming it's already in).

Django (#27)

• 22 queries covering ORM, migrations, URL routing, HTTP, auth,
  sessions, middleware, forms, templates, generic views, testing,
  signals, management commands
• 92.8% retrieval savings, 82.9% compression, 98.8% combined
• recall@10 = 0.95 (21/22 full-or-partial hits)

Express (#28)

• 20 queries covering app init, middleware, routing, request/response
  methods, views, utils
• 93.6% retrieval savings, 54.3% compression, 97.1% combined
• recall@10 = 1.00 (20/20 hits)

Reproduce

python benchmarks/run_benchmark.py --repo https://github.com/django/django.git --source-dir django
python benchmarks/run_benchmark.py --repo https://github.com/expressjs/express.git --source-dir lib

Test plan

• Both commands run end-to-end against a fresh clone
• Verified all previously-failing Django queries against real source
  identifiers (class/method names checked in the actual repo, not
  guessed) before finalizing wording