[clang] [llvm] Disable pre-link LTO vectorisation and loop unroll for at O2 and O3 optimisation levels (PR #195886)

[Luke Lau via cfe-commits]

lukel97 wrote:

> > The combine itself is missed because of multiple splats, identical in value, but different in SSA name.
> 
> Are we missing a run of CSE somewhere?

I think I post a fix for this in VPlan CSE soon, it looks relatively straightforward. We just need to teach it to handle single scalar loads. I don't think it needs to block this PR.

https://github.com/llvm/llvm-project/pull/195886