Performance Baselines (micro + macro)

Goal

Measurement section (no grind). Deliverables 01.1-01.3 already landed (bench workloads in compiler/oric/benches/lexer.rs + parser.rs, plans/parser-perf/baselines.json, Criterion parser-perf-v0) -> verify-first re-detects. 01.4 adds the macro anchor.

Work Items

• Run micro-benchmark baselines (lexer_core / lexer / parser raw throughput); confirm cross-run variance < 5%.
• Author + run new micro-workloads: string-literal-heavy lexer (generate_string_heavy/bench_string_heavy), expr-heavy parser (generate_expr_heavy/bench_expr_heavy), Salsa-overhead isolation (bench_salsa_overhead).
• Record baselines.json (micro anchors) and save the Criterion baseline parser-perf-v0 across lexer_core/lexer/parser for regression deltas.
• Establish a macro-benchmark baseline: end-to-end frontend throughput (lex+parse) on a realistic multi-file corpus; record it as the macro anchor.