Add benchmark for tree-sitter parsing speed across languages

**Scope:** `crates/scalpel-analyze/benches/` (new file)

Add criterion benchmarks measuring lines-per-second parsing speed for each supported language. Use sample files of ~1000 lines each.

**How to test:** `cargo bench -p scalpel-analyze`

**Difficulty:** Easy — criterion setup with sample files.