feat: add psutil comparison tests for cpu, mem and load

[shirou]

Summary

Compare gopsutil results against Python psutil — the project gopsutil was originally modeled after, and which we continue to look up to as the reference implementation — to catch parser and semantics regressions with real values.

We have already been using this approach in the process package since #1488 (TestName_Against_Python). This PR expands it toward all packages, but is just the foundation: a shared helper in internal/common/psutiltest plus the first batch of comparisons, covering cpu, mem and load:

• Eval[T](t, expr) runs python -c with psutil pre-imported and JSON-decodes the result (namedtuples → structs). Tests are skipped when python/psutil is not available; setting GOPSUTIL_PSUTIL_TEST turns skips into failures.
• Monotonic counters (cpu times, swap sin/sout, ctx switches) are asserted with a bracket: psutil is sampled before and after the gopsutil call and the gopsutil value must fall in between.
• Fluctuating gauges (available memory, load average) are resampled via require.EventuallyWithT with a relative tolerance plus an absolute floor.

All test names end in _Against_Psutil (go test -run Against_Psutil ./...).

Deliberately not compared (documented in code)

• VirtualMemoryStat.Used/UsedPercent: gopsutil uses Total - Available, psutil uses total - free - buffers - cached.
• Percent()-style instantaneous values.
• Fields that only exist in linux's /proc/stat accounting are gated to linux; darwin Free differs by design (psutil subtracts speculative pages).
• Swap comparison is linux-only for now (windows/darwin semantics unverified).

CI

test.yml installs python and a pinned psutil (the version these comparisons were verified against) on all runners, with continue-on-error so an install failure only skips the comparison tests. Once this proves stable, strict mode (GOPSUTIL_PSUTIL_TEST: require) can be enabled to make silent skips impossible.

Follow-ups planned: host, disk, net, and extending the process comparisons.

[Copilot]

Pull request overview

This PR introduces a shared test helper for comparing gopsutil outputs against Python psutil (as a reference implementation), and adds initial cross-implementation comparison tests for CPU, memory, and load metrics. It also updates CI to best-effort install Python + a pinned psutil version so these comparison tests can run in GitHub Actions.

Changes:

• Add internal/common/psutiltest helper package to evaluate psutil expressions via python -c, JSON-decode results, and provide comparison assertions/tolerances.
• Add new _Against_Psutil comparison tests for cpu, mem, and load, including bracketed assertions for monotonic counters and eventual/tolerant assertions for fluctuating gauges.
• Update the main test workflow to install Python and a pinned psutil version (best-effort) to enable these tests in CI.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
internal/common/psutiltest/psutiltest.go	Adds python/psutil discovery + Eval[T] helper with strict-mode support via env vars.
internal/common/psutiltest/assert.go	Adds bracketed and tolerance-based assertion helpers for stable comparison tests.
internal/common/psutiltest/psutiltest_test.go	Adds unit tests validating Eval behavior and error propagation.
cpu/cpu_psutil_test.go	Adds psutil comparison tests for CPU counts and CPU times (including per-CPU).
mem/mem_psutil_test.go	Adds psutil comparison tests for virtual memory and (linux-only) swap behavior.
load/load_psutil_test.go	Adds psutil comparison tests for load averages and context switch counters.
.github/workflows/test.yml	Installs Python + pinned psutil (best-effort) before running go test to enable comparisons in CI.

---

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