Fix SITL system load metric permanently pinned at 100%, blocking arming

[sensei-hacker]

Summary

A freshly-defaulted SITL build on 9.1 cannot arm: the Configurator's Status
tab shows CPU load pinned at/above 100%, tripping
ARMING_DISABLED_SYSTEM_OVERLOADED on an idle, correctly functioning
instance.

averageSystemLoadPercent (src/main/scheduler/scheduler.c) is a duty-cycle
estimator that counts how many scheduler() samples found a task already
overdue. That only produces a sane result if scheduler() runs far more
often than tasks become due — true on real hardware, which busy-polls. PR
#11436 made SITL's scheduler() sleep until just before the next task is
due instead of busy-polling (a real, worthwhile CPU-usage improvement on the
host). That eliminated the zero-load samples the old estimator relied on, so
every sample now finds >=1 task due and the average is permanently >=100.

Changes

• Cherry-picked 5e47ffd356 ("Fix SITL system load accounting") by
  @xznhj8129, originally authored against maintenance-10.x as part of
  SITL & X-Plane simulation fixes #11712. Full credit for the fix
  design and implementation goes to that commit's author — this PR ports it
  to release/9.1, where the same regression (8a475617c4, "SITL Tweaks")
  is also present.
  • Adds a SITL_BUILD-gated busy-time/elapsed-time duty-cycle calculation:
    scheduler() measures wall-clock time spent per iteration (excluding
    the intentional sleep), and taskSystem() converts that into a percentage
    over a rolling window, capped at 100.
  • Hardware (non-SITL_BUILD) load accounting is untouched.
  • docs/SITL/SITL.md: one-line note on how SITL load is now calculated.
• Follow-up commit (mine): moved a stale UNUSED(currentTimeUs) call into
  the branch where it's actually still unused, and added a short comment
  explaining why SITL needs a different calculation than hardware — raised
  by code review, no behavior change.

Testing

• Built SITL from this branch; froze a synthetic stress test into taskGyro()
  temporarily (reverted before commit) to drive genuine load:
  • Idle: cpuLoad (MSP2_INAV_STATUS) steady at 3-4%, no
    ARMING_DISABLED_SYSTEM_OVERLOADED.
  • 3,000 busy-loop iterations/call: cpuLoad 10-12%.
  • 20,000 busy-loop iterations/call: cpuLoad stable 58-59%, metric tracks
    added load, still armable, no overload flag.
• Host CPU usage of the idle SITL.elf process measured two ways (ps pcpu and /proc/[pid]/stat tick deltas): steady 7.4-8% of one core,
  confirming PR [SITL] Maintenance and Improvements #11436's sleep-based scheduling is untouched (no
  busy-polling reintroduced).
• Compiled a hardware target (MATEKF722, F7) cleanly with no warnings —
  confirms the non-SITL_BUILD path is unaffected (all changes are inside
  #if defined(SITL_BUILD) blocks).

Code Review

Reviewed with the inav-code-review agent: approved, no critical or
important issues blocking merge. Two minor suggestions (explanatory comment,
misplaced UNUSED()) were addressed in the follow-up commit.

Fixes #11710

[qodo-code-review]

Qodo reviews are paused for this user.

Troubleshooting steps vary by plan Learn more →

On a Teams plan?
Reviews resume once this user has a paid seat and their Git account is linked in Qodo.
Link Git account →

Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center?
These require an Enterprise plan - Contact us
Contact us →

[github-actions]

Test firmware build ready — commit ca26d31

Download firmware for PR #11720

243 targets built. Find your board's .hex file by name on that page (e.g. MATEKF405SE.hex). Files are individually downloadable — no GitHub login required.

Development build for testing only. Use Full Chip Erase when flashing.