Expose watchdog reset reason

[jackwilsdon]

This PR adds a reason() method to rp235x_hal::watchdog::Watchdog, returning the reset reason from the watchdog peripheral.

Both the force and timer bits could technically be set simultaneously, but I've not been able to reproduce this.

One thing to note is that some ways to reboot the RP2350 give unexpected reasons:

Method	Reason
reboot(RebootKind::Normal, RebootArch::Normal)	Timeout
explicit_buy	Timeout

The official SDK avoids this confusion by not exposing the reason directly and instead providing watchdog_caused_reboot (which checks reason and boot type) and watchdog_enable_caused_reboot (which checks reason and a scratch register set by watchdog_enable).

[thejpster]

I wonder if we should do it how the Pico SDK does it, to avoid the confusion.

[jackwilsdon]

We can implement watchdog_caused_reboot quite easily but we'd need to set a value in the scratch register on enable to implement watchdog_enable_caused_reboot - is this something we'd want to do? Happy to implement it in this PR if so.

[thejpster]

I think that would be good. I'd we use a scratch that someone else is using, they can switch to one of yes others easily enough - or just call a different reboot function that doesn't set the register.

[thejpster]

I'm currently using a scratch register in my project to catch whether we've had a Core 0 restart (thanks to probe-rs) or a full system restart. If I detect the former, I use the watchdog to force the latter, so that Core 1 also gets reset. I'd have no problem changing that if required.