Fix FMMU and SM reset

[david-boles]

It appears that clearing all SMs and FMMUs on the Kistler 5074B with a single block write doesn't work: In the 6th and 8th packets in the attached "fails" capture you can see that the working counter is zero after attempting to write to Ado 0x600 and 0x800.

I noticed because it results in an incorrect FMMU length being configured, which then causes the subdevice to clobber the PDI with garbage:

ethercrab/src/subdevice/configuration.rs

Lines 459 to 460 in 4ca7390

	let fmmu_config = if fmmu_config.enable {
	fmmu_config.length_bytes += sm_config.length_bytes;

I believe the lengths of the memory regions to blank were also wrong/off by one (16*16=256=0x100 for FMMUs, 16*8=128=0x80 for SMs).

Instead, I tried blanking each individually, and in the "succeeds" capture you can see that the first 8 SMs and 8 FMMUs are cleared successfully (working counter of 1); presumably the number of each that this device has.
sm_and_fmmu_reset.zip

[jamwaffles]

Very well spotted, and thanks for the fix! I recaptured the replay tests and pushed them to your branch to get the CI working again - hope you don't mind.