[clang] [llvm] [RFC] Use pre-compiled headers to speed up LLVM build (~1.5-2x) (PR #173868)

[Alexis Engelke via cfe-commits]

aengelke wrote:

FWIW, statistics about optimization passes >5s (no adaptors, wrappers, etc.) on libLLVM (NB: pass times can include analyses):
```
   Total   NumRuns       PerRun  Pass
   5.01s    160476      31.19us  LoopUnrollPass
   5.35s   4404446       1.21us  TargetIRAnalysis
   6.94s   4403527       1.58us  AssumptionAnalysis
   7.10s    433061      16.41us  LoopRotatePass
   7.91s      3909    2023.59us  CallGraphAnalysis
   9.66s      1980    4877.52us  CalledValuePropagationPass
  10.19s      1980    5145.05us  AlwaysInlinerPass
  12.22s   5587932       2.19us  PostDominatorTreeAnalysis
  12.26s   2070891       5.92us  ReassociatePass
  13.61s   2070891       6.57us  ADCEPass
  13.87s    737972      18.80us  LICMPass
  13.90s   2070891       6.71us  BDCEPass
  15.25s    432540      35.25us  LoopDeletionPass
  15.41s   4462734       3.45us  LoopSimplifyPass
  17.77s   5938233       2.99us  LoopAnalysis
  18.65s      3960    4710.34us  GlobalOptPass
  19.69s  11044561       1.78us  DominatorTreeAnalysis
  20.63s    302139      68.27us  IndVarSimplifyPass
  23.03s   2070891      11.12us  MemCpyOptPass
  25.45s    302139      84.23us  LoopIdiomRecognizePass
  25.85s   4139632       6.24us  PostOrderFunctionAttrsPass
  28.08s   2070891      13.56us  SCCPPass
  30.01s      1980   15158.71us  IPSCCPPass
  34.19s   3495375       9.78us  BranchProbabilityAnalysis
  42.01s   2843306      14.77us  MemorySSAAnalysis
  49.53s   3495375      14.17us  BlockFrequencyAnalysis
  64.37s   6536108       9.85us  SROAPass
  82.56s   4304741      19.18us  EarlyCSEPass
  91.52s  13070009       7.00us  SimplifyCFGPass
 248.60s   2069816     120.11us  InlinerPass
 279.41s  10996635      25.41us  InstCombinePass
1274.61s      1980  643743.37us  Optimizer

```

https://github.com/llvm/llvm-project/pull/173868