Fixed the RP2040 interrupt problem

[Snipeye]

I noticed a bug in the pulseio module for the RP2040 when using it:

Under specific circumstances, an item getting pushed into the queue during the PIO interrupt handler can be overwritten.

This happens specifically when the "len--" of popleft() is being executed: len is loaded into a register, then the register is decremented, then written back to memory. If the interrupt happens after len is loaded from memory into a register, or after it's decremented - but before it's written back to memory - then the value added in that interrupt will be overwritten (since the register state is restored when the interrupt ends, and the old decremented value of len is written back in, even though len was incremented in the handler).

The fix is not particularly elegant, but it requires no spinlocks or interrupt disabling, and in testing has not failed.

[dhalbert]

Thanks for coming up with an elegant fix. But it also perfectly ok to disable interrupts, and that's typically what we'd do in this kind of case, with common_hal_mcu_disable_interrupts() and common_hal_mcu_enable_interrupts(). That is straightforward, and fast to execute. It would be more robust against future code changes.

[dhalbert]

Thank you for noticing and fixing!