x86_64: Investigate performance regressions related to ASLR

[mkannwischer]

Out x86_64 backend performance seems to be impacted by stack placement.
See #1007 (comment) and also #1091.

We should investigate that more.

[mkannwischer]

• I have just rebased Bench: Disable ASLR on x86_64 to stabilize cycle counts #1091: Good news is that x86_64: 32-byte align Keccak x4 AVX2 stack frame #1124 seems to have solved the issue on c7i, c7a, c6a. However, ASLR still seems to have a large effect on c6i performance, so there must be something else going on here.