[clang] [llvm] [clang][OpenMP] Improve loop structure for distributed loops (PR #201670)

[Robert Imschweiler via cfe-commits]

ro-i wrote:

btw, sometimes it's a little hard to define what the MB in MB/s should be for a given reduction. But since the formula is consistent, it doesn't matter for the relative comparison if different people would define the MB constant differently.

with m = the MB constant, x = this PR's time (faster, larger MB/s), and y = base time:
```
((m/x - m/y) / (m/y)) * 100
  = (m/x)/(m/y) * 100 - 100
  = (y/x) * 100 - 100
  = ((y - x) / x) * 100
```

https://github.com/llvm/llvm-project/pull/201670