More robust option bit checking on STM32H7 targets

[labbott]

We recently discovered that our attempt at adding debugging for a flash issue #1908 was actually incorrect and prevented us from applying the option bit changes. We could be doing some more checks in https://github.com/oxidecomputer/hubris/blob/master/drv/stm32h7-update-server/src/main.rs#L220 (fn unlock) including checking that the bits actually wrote and checking FLASH_OPTSR.LOCK separately.