Rename pip_parse to pip_install

[groodt]

This is a fairly large refactor and unfortunately some breaking changes. The intent here is to reduce complexity and duplication.

Historically, we had 2 rules to install packages from PyPI (pip_install and pip_parse), with pip_install being the original. It worked well, but had a significant flaw in that it would eagerly fetch and install all packages. This was slow and frustrating when there were a large number of dependencies, so pip_parse was born!

pip_parse (even though it is still a repository rule) lazily fetched and installed packages. This was adopted by most of the community based on evidence from issues raised, the bazel Slack channels and the rule maintainers.

Maintaining both versions should no longer be necessary. There was a lot of duplication in both versions and it required extra effort and consideration for adding any new features.

I've taken the approach (probably 🌶️ ) that pip_install is a better name, so I've introduced this change where both pip_install and pip_parse share the same code, but with a small shim to handle the requirements_lock vs requirements difference in argument. I think a deprecation log message should be added and then pip_parse should be removed.

2 Breaking Changes:

1. The small breaking change in this PR, will be for the users (probably a small number) who still use a pip_install version prior to this PR and who DO NOT use the requirement() macro to reference dependencies. The naming style of the external repositories has changed, so this will require a sed across any .bazel files if these users upgrade. This isn't a big deal in my opinion.

2. A bigger breaking change might be this: https://github.com/bazelbuild/rules_python/pull/433 I'm not really sure this use case should be supported. If the code is available in the repo already, just use a py_library. If it's a "wheelhouse" or ".whl" file checked-in to the repo, Im sure we can wrangle a whl_install rule to unpack it more cleanly than using a requirements file.