Performance parity sprint: 3-cycle validation + Linux baseline hardening

[iahncajigas]

Summary

This PR delivers the post-merge performance parity sprint on codex/perf-parity-linux-sprint-20260304, including robustness hardening and repeated full-cycle validation.

Required Cycle Commits

• 520e431 — Cycle 1: tighten perf regression env checks and record full gate pass
• c05cc66 — Cycle 2: lock Linux perf baseline and optimize decoding hotspot
• d90ce06 — Cycle 3: full acceptance rerun green with no new fixes

Cycle Validation Checklist

• parity/CYCLE_VALIDATION_CHECKLIST.md

Decoding Optimization + Linux Baseline (Short)

• Optimized DecodingAlgorithms._compute_spike_rate_cis_matlab by removing the inner Monte Carlo draw loop bottleneck and vectorizing stimulus/history/probability computations while preserving MATLAB-parity behavior.
• Added/locked Linux performance baseline fixtures:
  • tests/performance/fixtures/python/performance_baseline_linux_20260304.json
  • tests/performance/fixtures/python/performance_baseline_linux_20260304.csv
• Expanded deterministic benchmark coverage with nspikeTrain.getSigRep and Analysis.fitGLM pipeline workloads.

CI Performance-Parity Baseline Confirmation

Confirmed: .github/workflows/performance-parity.yml now uses the Linux baseline fixture on ubuntu-latest and enforces strict regression environment matching:

• --previous-python-report tests/performance/fixtures/python/performance_baseline_linux_20260304.json
• --require-regression-env-match