Use NewPM wherever possible

[pchintalapudi]

Anywhere we fail to convert to NewPM/it's very tedious points to a hole in the NewPM API, so this serves to flush out remaining API issues.

Known issues:

• running a single NewPM pass requires a lot of boilerplate
• specifying alias analysis stack/target analyses is very tedious for short pipelines
• adding PassInstrumentationCallbacks to every run of a pass manager is also tedious

[codecov]

Codecov Report

Patch coverage: 88.02% and project coverage change: -7.90% ⚠️

Comparison is base (0b08f8c) 82.92% compared to head (bfff46d) 75.03%.
Report is 1 commits behind head on master.

❗ Current head bfff46d differs from pull request most recent head 38e1d38. Consider uploading reports for the commit 38e1d38 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #482      +/-   ##
==========================================
- Coverage   82.92%   75.03%   -7.90%     
==========================================
  Files          23       23              
  Lines        3069     3256     +187     
==========================================
- Hits         2545     2443     -102     
- Misses        524      813     +289     

Files Changed	Coverage Δ
src/GPUCompiler.jl	100.00% <ø> (ø)
src/irgen.jl	80.49% <ø> (-12.72%)	⬇️
src/driver.jl	65.28% <50.00%> (-27.11%)	⬇️
src/metal.jl	68.01% <50.00%> (-0.65%)	⬇️
src/optim.jl	86.91% <98.66%> (-0.19%)	⬇️
src/gcn.jl	67.61% <100.00%> (-0.61%)	⬇️
src/mcgen.jl	91.30% <100.00%> (-8.70%)	⬇️
src/ptx.jl	70.10% <100.00%> (-28.22%)	⬇️
src/rtlib.jl	95.00% <100.00%> (+0.12%)	⬆️
src/spirv.jl	82.39% <100.00%> (-10.71%)	⬇️

... and 9 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[maleadt]

Thanks @pchintalapudi for the initial version and the LLVM.jl updates!