fan: latch HBR tach above max_rpm/2 to absorb soft-PWM transients

[DRracer]

Summary

• New always_emit_tach bool property on the fan device. When set, fan_tach_expire ignores tach_blocked and fan_pwm_change latches current_rpm at max_rpm once it has been above max_rpm/2.
• Enabled only on fans[FAN_HBR] in prusa-mk4.c (print fan behaviour unchanged).
• The standalone Fan Test still passes both phases because the print-fan-test phase runs with HBR PWM=0 (no pulses emitted regardless of the latch), and the HBR-100% phase reads cfg.f_rpms[1] = 7500, inside the FW's [6800, 8700] acceptance window.

Motivation

On the MK4 board, PF13 is multiplexed between the HBR and print fan tach lines (HBR.tach-disable <- PF13, PRINT.tach-disable <- PF13 inverted in prusa-mk4.c:920-934) and is also bit-banged by the at21csxx loveboard EEPROM. The firmware's Fans::tick() toggles PF13 at 1 kHz to scan both fans; on real silicon the mux is a passive input read by a glue chip, but in QEMU there is no HiZ — the GPIO holds whatever the FW last drove — so the sim's tach_blocked path masks the HBR tach roughly half the time during normal operation.

This is invisible during the standalone fan selftest (the FW puts the fan into selftest_mode, which short-circuits CFanCtlCommon::is_fan_ok()), but after the heater test, autoFan re-engages and the software_pwm device's 256-tick window averaging yields small-but-nonzero intermediate values during the PWM=0 → PWM=255 transition. is_fan_ok() treats zero RPM as OK (short-circuit) but small RPM as a failure when commanded PWM is up, firing WarningType::HotendFanError ("Hotend fan not spinning") about 25 s into the cool-down — which kills the rest of the selftest snake.

The proper long-term fix is probably to honour STM32 GPIOx_MODER bits so the mux-select GPIO models HiZ correctly. This latch is a targeted workaround that unblocks the full first-boot selftest in sim.

Test plan

• Build clean on top of MINI404.
• Standalone Fan Test still passes on prusa-mk4-027c with mk4_release_noboot.bin (Buddy 6.6.0+14739): Print 6552/2658 RPM in range, Heatbreak 7500 RPM in range at both 100% and 40% PWM phases.
• No spurious HotendFanError after the heater test on the same FW build.
• Full first-boot selftest snake runs to completion (Fans → XYCheck → ZAlign → Loadcell → ZCheck → Heaters → Gears [skipped] → FS calibration → home), exercised end-to-end via a harness walker. No set_warning(0) in the CDC log.