Remove big-endian swizzles from vreinterpret

[sayantn]

Fixes: #1904

This is the correct behavior according to the Intrinsics Guide, which specifies that these should compile to NOP. Also, GCC and Clang matches this behavior.

I am a bit surprised that this didn't get caught in the AArch64_BE intrinsic_test run!

[rustbot]

r? @folkertdev

rustbot has assigned @folkertdev.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[folkertdev]

cc @Jamesbarford @adamgemmell

[adamgemmell]

I'm also really confused why the intrinsic-test tool doesn't catch this. Printing out the values you can see that the intrinsic does indeed get tested and it matches C. At the same time this PR clearly changes their behaviour but still passes the test tool

[folkertdev]

I've confirmed locally that the rust intrinsic-test-programs binary emits the same output on master and on this branch, but also that the generated assembly is in fact different (as expected based on the source code). So, eh, weird?

Just confirming with the Arm folks: the implementation in this PR is the correct one, right?

[Jamesbarford]

I've confirmed locally that the rust intrinsic-test-programs binary emits the same output on master and on this branch, but also that the generated assembly is in fact different (as expected based on the source code). So, eh, weird?

Just confirming with the Arm folks: the implementation in this PR is the correct one, right?

Looking through how it is implemented in arm_neon.h, there is no shuffling. This, to me, looks to be correct