Finalize `cortex-m-rtic-trace`

[tmplt]

cortex-m-rtic-trace must be finalized before release. The following must be done:

• [x] Take user-supplied information (TPIU parameters, GTS generation, output protocol, etc). We should probably define some Configuration struct.
• [x] Feature availability check: if the hardware doesn't support the given configuration an error should be returned.
• [x] The stm32f4 specific code should be removed. The crate should only handle core peripherals. Device-specific sequences should be done inline (or sent upstream into probe-rs).
• [x] The watch variables should be u8 instead of u32: an RTIC app is unlikely to have more than 255 software tasks, and making them u8 simplifies some code in mod recovery.

Unfortunately, I do not think we can close #88.

[tmplt]

Something else to consider: the two watch variables are adjacent to each other. If they end up on the same page in memory, could we possibly end up with spurious events? What about user-code variables? Need we completely isolate the variables from other code, but also from eachother?