interrupts: efficient user interrupt event push

[pthierry-ledger]

By now, the kernel delivers one interrupt at a time to a userspace task that is requesting waiting interrupt from its queue using the sys_waitforevent() syscall.
Instead, the kernel should be able to delivers at once the complete list of interrupts so that the userspace thread can:

• get back all interrupt at once with a single system call
• decide in which order all waiting IRQ can be handled (userspace-local priority decision)

This will highly increase interrupt efficiency & reactivity for projects that manupulate (a lot of-) interrupts. This mostly requires
that the waiting IRQ vector should be pushed to userspace through SVC exhange area, and thus requiring a header that specify the length of the transmitted IRQ vector.