fix(benchmark): run typedframes checker against actual corpus

[w-martin]

Summary

• Fixes fix(benchmark): typedframes binary always targets examples/ not the benchmark corpus #1: switches typedframes to use the Python CLI (uv run typedframes check) so the benchmark targets the actual corpus directory rather than a single hardcoded example file
• Removes the dead find_binary helper and binary_path parameter
• Removes the invalid "10-100x faster than mypy" claim
• Re-runs the benchmark against both corpora with the corrected code; the great_expectations column now shows the real figure (310ms ±10ms, not the bogus 930µs)
• Clones great_expectations to a tempfile.mkdtemp() dir and removes it after copying, so no large directory is left behind in /tmp
• Omits tools from the README table if they couldn't be found at benchmark time (e.g. pyright not installed locally)

Stacked on #18 (cache-clearing path fixes) — target that branch, not main.

Test plan

• run_codebase_benchmarks no longer special-cases typedframes
• typedframes uses uv run typedframes check as its command
• great_expectations clone is cleaned up after the benchmark
• README updated with corrected benchmark figures
• Tools unavailable locally are omitted from the table (not shown as "Tool not found")

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!