feat: 60 Hz frame-rate support — internal FSIN rate select + frame-budget scaling

[boringethan]

Summary

Support 60 Hz frame rate (16.7 ms frame period vs 25 ms at 40 Hz). External-FSIN (laser-on) capture already follows whatever rate the console trigger produces; the gaps are:

1. Internal FSIN (laser-off) is hardcoded 40 Hz — TIM4 PSC=1000/ARR=5999 in MX_TIM4_Init. Add a rate parameter to OW_CAMERA_FSIN (reserved field: 0=off, 1=on@40 for back-compat, 40/60=explicit rate) and reconfigure ARR accordingly.
2. FSIN debounce is 15 ms (camera_manager.c ~1516) — at 60 Hz the frame spacing is 16.7 ms, leaving <2 ms jitter margin before real frames get swallowed. Reduce to 10 ms (still generous for ringing).
3. Compression budget guard assumes 25 ms frames — CMP_BUDGET_HARD_US 15 ms + copy/send/re-arm does not fit a 16.7 ms frame. Scale hard/warning budgets to be safe at 60 Hz (fallback-to-uncompressed is the safe path).

IMU (TIM14) stays at 40 Hz sampling for now — histograms and IMU stream independently; noted as a possible follow-up.

Part of the cross-repo 60 Hz capture mode (bloodflow-app umbrella issue).

Validation

Bench: 60 Hz external-FSIN scan on both sensors (16 cams), verify ~60 fps arrival, zero SPI overruns / camera-death false positives; laser-off internal-FSIN at 60 Hz.

[boringethan]

Implemented on feature/78-60hz-support (3b3a089, Debug build clean at 37% flash): OW_CAMERA_FSIN reserved now carries the rate (0=off, 1=legacy 40, 40/60 explicit) via new X02C1B_fsin_set_rate(); FSIN debounce 15→10 ms; rle_compress budgets 15/10 → 11/7 ms for the 16.7 ms worst-case frame. External-FSIN 60 Hz needs no other firmware change — bench validation queued (waiting for the rig to free up).

[boringethan]

Bench validation results (2026-07-04 night, both sensors, all 16 cameras, Debug build from this branch)

40 Hz clinical regression — clean. 20 s laser-on scan: 16/16 cameras at exactly 40.00 Hz median inter-frame, 809 frames each, zero frame-id gaps, 2 darks each on the 600-frame schedule, light means 166–246 DN (pedestal ~66), no interlock faults, no debounce/budget regressions. (investigations/60hz_capture_mode/run9_40on.log)

60 Hz — blocked by the camera sensor, not by this firmware. With this build flashed, a 60 Hz external-FSIN scan delivers exactly one frame per camera then locks up. Knife-edge testing pinned the ceiling at the OV2312's configured frame time (25.0 ms → 40 fps): 39.5 Hz streams clean, 40.5 Hz locks up. Filed #80 for the camera timing mode (HTS/PLL) work; this branch's changes (rate-select command, debounce, budgets) are validated as not regressing 40 Hz and stand ready for when #80 lands.

Also validated on old firmware for comparison: 60 Hz there delivered ~2 batches before the 15 ms debounce + failure detector killed all cameras — confirming this branch's 10 ms debounce motivation, though the deeper camera-timing wall sits behind it.