drivers: rtc: rtc_ll_stm32: replace k_mutex with k_spinlock

[jeppenodgaard]

This allows using this RTC driver from contexts where k_mutex use is not allowed.

Rework rtc_stm32_set_time to lock just before and just after accessing functions that modifies register values.

[mathieuchopstm]

To avoid functional changes, the following should be used instead:

Suggested change

	k_spinlock_key_t key = k_spin_lock(&data->lock);
	k_spinlock_key_t key;
	int err = k_spin_trylock(&data->lock, &key);
	if (err < 0) {
	return err;
	}

cc @erwango for whether the functional change is OK or not

[jeppenodgaard]

For reference. These set_time functions use k_spin_lock:

• rtc_sam0_set_time
• rtc_numaker_set_time
• rtc_mc146818_set_time
• ifx_cat1_rtc_set_time
• ambiq_rtc_set_time
• rtc_rpi_pico_set_time
• ds1307_set_time

k_spin_trylock is not used by any rtc driver atm.

[jeppenodgaard]

@erwango any chance you got time to give this a look?

[erwango]

I don't think there's a good response here.
I agree you @mathieuchopstm, you're changing the dynamic here, but drivers behaves today differently from the other drivers indeed, so we can't argue about we'll fail on the API contract.
I'm fine with keeping this change and if some applications have trouble with that, let's say that we're moving to the general API consensus.

[mathieuchopstm]

Same remark here

[erwango]

@jeppenodgaard Rebase required