Doing so can cause deadlocks or panics during boot. In QEMU I was
seeing occasional panics during boot when a statclock interrupt
interrupted an idle thread in init_secondary(), before per-CPU scheduler
data structures were linked. It also seemed to be causing some
mysterious boot failures in AWS but I don't have console access and so
cannot confirm.
The idle threads are set up to release a spinlock section and enable
interrupts in fork_exit(), so there is no need to enable interrupts
earlier.