Bug: Spurious LAPIC EOI caused by non-LAPIC interrupts
I suppose we should leave it to interrupt handlers themselves to call LAPIC_EOI? Works fine as long as we have only one handler installed on the vector number.