ci: Add PR benchmark comparison workflow

[mrecachinas]

Summary

This PR adds a new GitHub Actions workflow that automatically compares benchmark performance between PR branches and the base branch.

Features

• Parallel benchmarking: Runs benchmarks on both base and PR branches simultaneously
• Detailed comparison: Calculates percentage change for each benchmark
• PR comments: Posts/updates a formatted table showing:
  • Benchmark name
  • Base branch time (median)
  • PR branch time (median)
  • Delta (% faster/slower)
  • Status indicator
• Status indicators:
  • ✅ Faster (>5% improvement)
  • ➖ Within noise (±5%)
  • ⚠️ Slower (5-20% regression)
  • ❌ Severe regression (>20% slower)
• Automatic failure: Workflow fails if any benchmark has >20% regression
• Comment updates: Uses peter-evans/find-comment and create-or-update-comment to update existing comments instead of creating new ones

Technical Details

• Uses ubuntu-latest runner with Python 3.11
• Benchmarks run with --benchmark-disable-gc for consistent results
• Uses median time (more stable than mean) for comparisons
• Inline Python script for comparison logic (no external dependencies)

[github-actions]

📊 Benchmark Comparison Results

Status	Benchmark	Base	PR	Delta
✅	bytes_arr_within[ 1024 elems,s=32,end]	0.0223ms	0.0196ms	12.4% faster
✅	bytes_arr_within[ 1024 elems,s=32,mid]	0.0112ms	0.0099ms	11.6% faster

➖ 22 benchmarks within noise (click to expand)

Benchmark	Base	PR	Delta
bytes_arr_within[ 512 elems,s=16,at 0]	0.0001ms	0.0001ms	1.7% faster
bytes_arr_within[ 512 elems,s=16,end]	0.0018ms	0.0018ms	0.7% faster
bytes_arr_within[ 512 elems,s=16,mid]	0.0010ms	0.0010ms	0.3% faster
bytes_arr_within[ 1024 elems,s=32,at 0]	0.0001ms	0.0001ms	1.5% faster
bytes_arr_within[16384 elems,s=64,at 0]	0.0001ms	0.0001ms	2.2% faster
bytes_arr_within[16384 elems,s=64,end]	0.0926ms	0.0854ms	7.7% faster
bytes_arr_within[16384 elems,s=64,mid]	0.0460ms	0.0428ms	7.0% faster
hex_within_dist	0.0002ms	0.0002ms	13.6% slower
hd_bytes[1000-diff]	0.0002ms	0.0002ms	0.1% slower
hd_bytes[1000-same]	0.0002ms	0.0002ms	0.3% slower
hd_bytes[1024-diff]	0.0002ms	0.0001ms	2.7% faster
hd_bytes[1024-same]	0.0001ms	0.0001ms	2.1% faster
hd_bytes[3-diff]	0.0001ms	0.0001ms	2.6% faster
hd_bytes[3-same]	0.0001ms	0.0001ms	2.9% faster
hd_bytes[64-diff]	0.0001ms	0.0001ms	0.6% faster
hd_string[1000-diff]	0.0003ms	0.0003ms	0.6% slower
hd_string[1000-same]	0.0003ms	0.0003ms	0.8% slower
hd_string[1024-diff]	0.0003ms	0.0003ms	2.5% slower
hd_string[1024-same]	0.0003ms	0.0003ms	0.9% slower
hd_string[3-diff]	0.0002ms	0.0001ms	41.4% faster
hd_string[3-same]	0.0001ms	0.0001ms	0.6% faster
hd_string[64-diff]	0.0001ms	0.0001ms	5.2% slower

Legend: ✅ Faster (>5%) · ⚠️ Slower (5-30%) · ❌ Regression (>30%) · 🆕 New · 🗑️ Removed