[aarch64] Enable Always and Never Branch patching for TBZ/TBNZ and CBZ/CBNZ

[galenbwill]

Addresses #7280

This PR implements the fix, but someone should sign off on this with respect to the following concern:

while i can fix always branch (by forcing it to be a TBZ/CBZ and changing the tested register to the zero XZR/WZR so the branch always happens), its’s not clear to me that it’s “ok”, in the sense that it changes the data dependency of the instruction, from the tested register to the zero register.

[ElykDeer]

I'm okay with the data-flow caveat; x86 will completely nop things out for some branches, and destroys register references in others as well. These changes look fine to me....but would there be any benefit to using a nop in some cases? Unless someone can think of a reason I think it's fine as-is.