build: Run benchmarks in parallel

[pjbgf]

Ensure the benchmarks are executed in parallel, so that they can finish within 30 mins even with count number of 6, which is the minimum for benchstat results.

[Copilot]

Pull request overview

This pull request restructures the benchmark regression workflow to execute base and PR branch benchmarks in parallel GitHub Actions jobs, reducing total execution time. The benchmark count is increased to 6 (required for benchstat statistical reliability) and file cleanup code is removed from benchmark functions since unique filenames are now used for each iteration and automatic cleanup via b.TempDir() handles directory cleanup.

Changes:

• Restructured GitHub Actions workflow to run base and PR benchmarks in parallel jobs
• Added environment variables for benchmark configuration (count=6, filter for memfs/osfs only)
• Removed per-iteration file cleanup code from benchCreate and benchWrite functions

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
.github/workflows/bench-regression.yml	Split single sequential benchmark job into three parallel jobs (setup, bench-base, bench-pr) plus a compare job; added environment variables for configuration; implemented artifact-based result sharing
test/bench_test.go	Removed per-iteration file cleanup from benchCreate and benchWrite benchmark functions to support parallel execution

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.