Fix MMU3 fsensor handling

[Lapeno89]

What I changed:

If the fsensor remains ON, the MMU3 attempts (up to two times) to retract 1 mm of filament.
If the fsensor turns OFF, the filament is fully unloaded.
If it stays ON, an fsensor failure is reported on the printer display.

[github-actions]

All values in bytes. Δ Delta to base

ΔFlash	ΔSRAM	Used Flash	Used SRAM	Free Flash	Free SRAM
168	8	28332	1681	340	879

[github-actions]

Automated Test Code Coverage Report

View details...

File	Lines	Exec	Cover
src/application.cpp	169	14	8%
src/application.h	3	3	100%
src/hal/circular_buffer.h	52	52	100%
src/hal/eeprom.h	1	0	0%
src/hal/gpio.h	18	7	38%
src/hal/progmem.h	6	6	100%
src/hal/tmc2130.cpp	113	9	7%
src/hal/tmc2130.h	31	27	87%
src/logic/command_base.cpp	139	118	84%
src/logic/command_base.h	4	3	75%
src/logic/cut_filament.cpp	117	80	68%
src/logic/eject_filament.cpp	77	60	77%
src/logic/feed_to_bondtech.cpp	60	57	95%
src/logic/feed_to_bondtech.h	1	1	100%
src/logic/feed_to_finda.cpp	48	45	93%
src/logic/feed_to_finda.h	1	1	100%
src/logic/home.cpp	18	12	66%
src/logic/load_filament.cpp	84	76	90%
src/logic/load_filament.h	1	0	0%
src/logic/move_selector.cpp	21	0	0%
src/logic/no_command.h	2	1	50%
src/logic/retract_from_finda.cpp	27	21	77%
src/logic/retract_from_finda.h	1	1	100%
src/logic/set_mode.cpp	5	0	0%
src/logic/set_mode.h	1	0	0%
src/logic/start_up.cpp	38	26	68%
src/logic/start_up.h	4	4	100%
src/logic/tool_change.cpp	107	85	79%
src/logic/unload_filament.cpp	81	60	74%
src/logic/unload_to_finda.cpp	77	52	67%
src/logic/unload_to_finda.h	1	1	100%
src/modules/axisunit.h	21	21	100%
src/modules/buttons.cpp	11	11	100%
src/modules/buttons.h	7	7	100%
src/modules/crc.h	13	13	100%
src/modules/debouncer.cpp	28	24	85%
src/modules/debouncer.h	7	7	100%
src/modules/finda.cpp	7	3	42%
src/modules/finda.h	2	2	100%
src/modules/fsensor.cpp	6	6	100%
src/modules/fsensor.h	3	3	100%
src/modules/globals.cpp	47	42	89%
src/modules/globals.h	34	24	70%
src/modules/idler.cpp	80	73	91%
src/modules/idler.h	12	12	100%
src/modules/leds.cpp	56	54	96%
src/modules/leds.h	16	15	93%
src/modules/math.h	6	6	100%
src/modules/motion.cpp	59	40	67%
src/modules/motion.h	66	64	96%
src/modules/movable_base.cpp	73	69	94%
src/modules/movable_base.h	16	16	100%
src/modules/permanent_storage.cpp	144	89	61%
src/modules/protocol.cpp	216	184	85%
src/modules/protocol.h	72	70	97%
src/modules/pulley.cpp	26	25	96%
src/modules/pulley.h	9	5	55%
src/modules/pulse_gen.cpp	95	89	93%
src/modules/pulse_gen.h	53	51	96%
src/modules/selector.cpp	69	62	89%
src/modules/selector.h	5	5	100%
src/modules/speed_table.h	26	24	92%
src/modules/user_input.cpp	39	39	100%
src/modules/user_input.h	12	12	100%
src/modules/voltage.cpp	4	0	0%
src/registers.cpp	86	22	25%
src/unit.h	12	12	100%
TOTAL	2746	2023	73%

TOTAL: 2746 lines of code, 2023 lines executed, 73% covered.

[gudnimg]

@Lapeno89 thanks for opening a PR. The PR in its current state is not ready for a merge.

1. The unit tests need to pass to ensure your change is not introducing regressions.

2. It would also help us reviewers if you would improve the PR description. Like what is the problem this PR is solving exactly (with as much detail as possible)? And how do I reproduce the issue this PR is aiming to fix.

The PR consumes 168 bytes of flash which is a lot when considering how much free flash memory is left. Not a blocker in and of itself, but before we can merge, it would be good to optimize the code as much as possible.