Speed up pipx list by removing multiprocessing code

[itsayellow]

• I have added an entry to docs/changelog.md

Summary of changes

Since we are not compute-limited in pipx list, parallel code is not appropriate, and in fact makes running pipx list noticeably slower than purely sequential code.

See #616 for some timing benchmarks

This PR removes the multiprocessing code for pipx list, and makes it always use fully sequential code.

Closes #616

Test plan

Tested by running

Unix / macOS;

> time pipx list

Windows:

> Measure-Command { pipx list }

[uranusjr]

Semi-related, it seems like multiprocessing was used because the implementation used to call pip list. This also left a superfulous python_path argument along the call stack that can be removed in a follow-up refactoring.

[uranusjr]

I’ve filed #626 to remove the unused variable.

[itsayellow]

Semi-related, it seems like multiprocessing was used because the implementation used to call pip list. This also left a superfulous python_path argument along the call stack that can be removed in a follow-up refactoring.

Ahh, that makes more sense. Thanks for going back in the history and finding that out.

[itsayellow]

This seems to be the actual commit when @cs01 made list_packages() parallel: af61325

And yes, it looks like functions run from the Venv.py module involved some subprocess calls. Making them parallel might've helped more then than it does now.