Feat/clock domain

[FoniksFox]

Closes #665

Add ClockDomain, a centralised clock validator and applicator

Clock configuration is a precomputed ClockTree (from a host tool or hand-written). The firmware validates it at compile time and applies it at runtime. No solver runs in the firmware.

• ClockTree stores only decisions (M/N/P/Q/R, prescalers, source enums); all frequencies are derived via accessors (sysclk(t), source_frequency(t, src), etc.)
• Board<> takes a mandatory ClockTree as the second template parameter
• Peripherals register clock requirements via ClockDomain::Device with typed models: SPIClockModel<Group, MaxBaud, MinBaud>, ADCClockModel<MaxADCCLK>, SDClockModel<MaxFreq, MinFreq>
• ClockDomain::build() validates PLL ranges, bus consistency, and peripheral requirements at compile time - refuses to compile on failure
• ClockDomain::Init::init(tree) applies the validated tree to HAL registers
• ADC prescaler computed from kernel clock at init (no longer user-specified)
• SPI/SD use mandatory frequency ranges
• Removed HALconfig

Clock-tree solver was not implement in compile time because of the added complexity (worst cases were not possible to compile unless willing to let the compiler do billions of operations). The solver may be implement at a further date via a brute force function in the domain that only exists on simulator builds, such that you can build your project with the simulation preset, solve the clock-tree and print it to a header file (it could be added as a build step or as a standalone script).

[github-actions]

ST-LIB Release Plan

• Current version: 6.1.3
• Pending changesets: 2
• Highest requested bump: major
• Next version if merged now: 7.0.0

Pending changes

• major Add ClockDomain, a centralised clock validator and applicator (.changesets/clock-redesign.md)
• patch Fix a warning about symbols in HardfaultTrace (doesn't change any behaviour) (.changesets/fix-hwarfault-warning.md)

[victor-Lopez25]

There is a lot of code duplicated here which shouldn't be, TimerDomain is not to be used like TimerWrapper. TimerWrapper already handles clocks, so remove TimerRequest from TimerDomain::Config and clock_frequency from the timer instance, as you can already get that from the timeridx or the TIM_TypeDef, it's fine to have a runtime check for if a timer is in apb1 or apb2 if you only have the TIM_TypeDef and that's exactly the reason why I have that function in TimerDomain.

[victor-Lopez25]

If you need the ClockDomain from the TimerDomain to know what apb1 and apb2 frequencies are, you should probably use a global ClockDomain, that would keep it simpler and you could even get the function which uses HAL code to check a TIM_TypeDef's frequency to work. The only place where it is used is in InputCapture, which is pretty important...

[victor-Lopez25]

This should just use a global:

if constexpr (this->is_on_APB1) {
  return ClockDomain::timer_apb1(global_clock_tree);
} else {
  return ClockDomain::timer_apb2(global_clock_tree);
}

You're also adding a PANIC where there's no reason to have one, a timer is always in apb1 or apb2 so it doesn't make much sense to add a possibility for failiure.

[victor-Lopez25]

Is there no way to have it work like the others, so we can simplify creating Board?

[victor-Lopez25]

If we already have a global ClockDomain instance, Board::clock_tree(), why are we passing the clock tree to every domain? I get that you can't access Board from a domain but you could just make it global from ClockDomain since you already know there will only ever be one.