Disable IRQs in OQueue spinlocks.

[arthurp]

This makes them safer to operate on inside IRQ handlers.

[ioeddk]

LGTM.
Additional comment: the SpinLock class says:

/// The guard behavior can be temporarily upgraded from [`PreemptDisabled`] to
/// [`LocalIrqDisabled`] using the [`disable_irq`] method.
///
/// [`disable_irq`]: Self::disable_irq

So if we don't want to disable IRQ in every OQueue, we can keep it preempt disabled only, and disable it ad hoc.

[arthurp]

LGTM. Additional comment: the SpinLock class says:

/// The guard behavior can be temporarily upgraded from [`PreemptDisabled`] to
/// [`LocalIrqDisabled`] using the [`disable_irq`] method.
///
/// [`disable_irq`]: Self::disable_irq

So if we don't want to disable IRQ in every OQueue, we can keep it preempt disabled only, and disable it ad hoc.

That would have been the "minimal" case I was thinking about. I think for any given OQueue either no access or all accesses need to disable IRQs, so ad-hoc usage is error prone. Maybe with the new OQueue we need a way to create "IRQ-safe" OQueues which do this disabling and normal ones which panic if accessed from an IRQ context.

I'm not going to merge this at the moment. It's definitely possible it breaks something and I don't want to make main a mine field without good reason. Maybe I'll merge later this week once we have a chance to chat about it.