Expose compression control for Pex zipping operations.

[jsirois]

The experiments here and here in #1675 prove out that the zip compression level provides a significant swing in tradeoffs. It's probably worth exposing this to let users decide where they want to land in the tradeoff spectrum.

[jsirois]

For a modest sized PEX the gain is modest, ~25% speed improvement creating the PEX zip, ~no runtime perf impact and ~350% size increase:

$ hyperfine -w1 'python -mpex jupyter -c jupyter -o jupyter.default.pex' 'python -mpex jupyter -c jupyter -o jupyter.no-compress.pex --no-compress'
Benchmark 1: python -mpex jupyter -c jupyter -o jupyter.default.pex
  Time (mean ± σ):      8.695 s ±  0.288 s    [User: 6.429 s, System: 0.253 s]
  Range (min … max):    8.422 s …  9.351 s    10 runs
 
Benchmark 2: python -mpex jupyter -c jupyter -o jupyter.no-compress.pex --no-compress
  Time (mean ± σ):      7.065 s ±  0.109 s    [User: 4.665 s, System: 0.276 s]
  Range (min … max):    6.928 s …  7.254 s    10 runs
 
Summary
  'python -mpex jupyter -c jupyter -o jupyter.no-compress.pex --no-compress' ran
    1.23 ± 0.05 times faster than 'python -mpex jupyter -c jupyter -o jupyter.default.pex'

$ ls -lh jupyter.*.pex
-rwxr-xr-x 1 jsirois jsirois 25M Apr  4 11:05 jupyter.default.pex
-rwxr-xr-x 1 jsirois jsirois 88M Apr  4 11:06 jupyter.no-compress.pex

$ hyperfine -w1 './jupyter.default.pex --version' './jupyter.no-compress.pex --version'
Benchmark 1: ./jupyter.default.pex --version
  Time (mean ± σ):     861.9 ms ±   6.8 ms    [User: 812.1 ms, System: 49.6 ms]
  Range (min … max):   852.6 ms … 873.5 ms    10 runs
 
Benchmark 2: ./jupyter.no-compress.pex --version
  Time (mean ± σ):     871.5 ms ±   9.5 ms    [User: 826.1 ms, System: 45.1 ms]
  Range (min … max):   859.8 ms … 885.8 ms    10 runs
 
Summary
  './jupyter.default.pex --version' ran
    1.01 ± 0.01 times faster than './jupyter.no-compress.pex --version'