perf(core): reduce allocations and memory in ProfileAnalyzer

[m2papierz]

What

Replace two per-bucket counter Vecs with a single Vec bitfield in ProfileAnalyzer::analyze, and add
unit tests for the cumulative helper methods on ExecutionProfile.

Why

Bottleneck classification only checks presence (> 0), never reads actual stall/failure counts. Tracking two
Vecs (8 bytes per bucket) for a boolean signal wastes memory and adds an unnecessary .zip() in the
post-loop pass. Per-stall detail is already preserved in stall_events, per-failure detail in the raw Trace.

How

• Bitfield replacement (analyzer.rs): stalls_in_bucket: Vec and factory_failure_counts: Vec replaced
  by bucket_flags: Vec with HAD_STALL = 1 and HAD_FAILURE = 2 flag bits. 75% memory reduction per bucket.
  Bottleneck classification iterates one Vec instead of zipping two.
• GateServed arm reorder (analyzer.rs): magic-state counting moved before the wait > 0 stall branch —
  unconditional work first, conditional second.
• Cumulative helper tests (profile.rs): 5 unit tests covering prefix-sum correctness, empty input, p_logical
  scaling, zero probability, and last-element-equals-total invariant.

Testing

• make ci passes locally (fmt + clippy + test + audit)
• New behavior has tests
• Hot-path changes have criterion benchmarks

Checklist

• PR description explains why, not just what
• No new unwrap()/expect() in production code
• No new allocations in the simulation hot loop
• Crate boundaries respected (pirx-core never imports from pirx-adapters)

[codspeed-hq]

Merging this PR will improve performance by 12.68%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 3 improved benchmarks
✅ 12 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	analyze[10]	13.8 µs	12 µs	+14.78%
⚡	analyze[500]	314.1 µs	280.9 µs	+11.82%
⚡	analyze[100]	71.6 µs	64.3 µs	+11.48%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing feat/core-perf (766e3c7) with master (f3cd97b)}