[lld] [lld][MachO] Parallelize parsing relocations for input files. (PR #223795)

[Liza Burakova via llvm-commits]

liza371 wrote:

Some extra benchmarks on different repros for reference

**mac_debug_component_sym_2**
```
Baseline 1
Time (mean ± σ):      7.425 s ±  0.041 s    [User: 7.359 s, System: 1.605 s]
Range (min … max):    7.321 s …  7.492 s    20 runs

Optimized
Time (mean ± σ):      7.116 s ±  0.087 s    [User: 7.640 s, System: 1.884 s]
Range (min … max):    6.998 s …  7.349 s    20 runs

Baseline 2
Time (mean ± σ):      7.360 s ±  0.079 s    [User: 7.285 s, System: 2.278 s]
Range (min … max):    7.269 s …  7.573 s    20 runs
```

**mac_asan_fat**
```
"Baseline 1
Time (mean ± σ):      8.151 s ±  0.062 s    [User: 7.688 s, System: 2.699 s]
Range (min … max):    8.047 s …  8.249 s    20 runs

Optimized
Time (mean ± σ):      7.629 s ±  0.082 s    [User: 8.325 s, System: 3.483 s]
Range (min … max):    7.459 s …  7.835 s    20 runs

Baseline 2
Time (mean ± σ):      8.176 s ±  0.036 s    [User: 7.712 s, System: 2.714 s]
Range (min … max):    8.082 s …  8.231 s    20 runs"
```

As a note: I realized when running `mac_debug_component_sym_2` that the baseline was off from my last PR's baseline, so I redid those and left a comment  [here](https://github.com/llvm/llvm-project/pull/222087#issuecomment-5687279946). I double checked activity monitor and my CPU was otherwise pretty quiet during all these runs and the results were pretty consistent, so I am more confident in my ability to properly run hyperfine now 💀 

https://github.com/llvm/llvm-project/pull/223795