This PR moves the parsing of `Requires-Dist` to the loading phase
within the `whl_library_targets_from_requires` macro. The original
`whl_library_targets` macro has been left unchanged so that I don't have
to reinvent the unit tests - it is well covered under tests.

Before this PR we had to wire the `target_platforms` via the
`experimental_target_platforms` attr in the `whl_library`, which means
that whenever this would change (e.g. the minor Python version changes),
the wheel would be re-extracted even though the final result may be the
same.

This refactor uncovered that the dependency graph creation was incorrect
if we had multiple target Python versions due to various heuristics that
this had. In hindsight I had them to make the generated `BUILD.bazel`
files more readable when the unit test coverage was not great. Now this
is unnecessary and since everything is happening in Starlark I thought
that having a simpler algorithm that does the right thing always is the
best way.

This also cleans up the code by removing left over TODO notes or code
that no longer make sense.

Work towards #260, #2319

(cherry picked from commit a19e1e4)

Also escape plus and percent when generating the repo name from the
wheel version.
Sometimes they have such characters in them.

Fixes #2799

Co-authored-by: Richard Levasseur <rlevasseur@google.com>
(cherry picked from commit e14cd37)

The wheel `METADATA` parsing implemented in 1.4 missed the fact
that whitespace is significant and sometimes License is included
inline in the `METADATA` file itself.

This change ensures that we stop parsing the `METADATA` file only
on first completely empty line.

Fixes #2796

---------

Co-authored-by: Ignas Anikevicius <240938+aignas@users.noreply.github.com>
(cherry picked from commit 1d69ad6)

)

Add `<micro>` version to the target platform. Instead of `cpxy_os_cpu`
the target platform string format becomes `cpxy.z_os_cpu`. This is a
temporary measure until we get a better API for defining target
platforms.

Summary:
- [x] test `select_whls` function needs to be tested to ensure that the
  whl selection is not impacted when we have the full version in the
  target platform.
- [ ] `download_only` legacy whl code path in `bzlmod` needs further
  testing.
- [x] test `whl_config_setting` handling and config setting creation.
  The config settings in the hub repo should not use the full version,
  because from the outside, the whl is compatible with all `micro`
  versions of a given `3.<minor_version>` of the Python interpreter.
  This means that the already documented config setting do not need to
  be changed.
- [x] `pep508_deps` tests for handling the `full_python_version`
  correctly.
- [x] `pep508_deps` tests for ensuring the `default_abi` is being
  handled correctly.

Fixes #2319

(cherry picked from commit 1e21dbd)

`bzlmod` has the full python version information statically and we don't
need to call Python to get its version, but for `WORKSPACE` that is not
the case and we have to call it before evaluating the markers in
universal requirements files.

This also fixes transitions in the `compile_pip_requirements` macro
where
the `.update` target would not transition correctly based on the
`python_version` parameter.

Fixes #2818

(cherry picked from commit ee34409)

Remove a stray file

(cherry picked from commit c981569)

This just adds the code back at the original state before the following
PRs have been made to remove them: #2629, #2781. This has not been
hooked up yet in `evaluate_markers` and `whl_library` yet and I'll need
extra PRs to do that.

No CHANGELOG entries for now, will be done once the integration is back.

Work towards #2830

(cherry picked from commit 61c91fe)

Handling of `python_full_version` correctly has been fixed in the
Starlark
implementation in #2793 and in this PR I am backporting the changes to
handle
the full python version target platform strings so that we can have the
same
behaviour for now.

At the same time I have simplified and got rid of the specialization
handling
in the Python algorithm just like I did in the starlark, which
simplifies the
tests and makes the algorithm more correct.

Summary:
* Handle `cp3x.y_os_arch` strings in the `platform.py`
* Produce correct strings when the `micro_version` is unset. Note, that
we use version `0` in evaluating but we use the default version in the
config setting. This is to keep compatibility with the current behaviour
when the target platform is not fully specified (which would be the case
for WORKSPACE users).
* Adjust the tests and the code to be more similar to the starlark impl.

Work towards #2830

(cherry picked from commit 9e613d5)

Summary:
- Revert to using Python for marker evaluation during parsing of
  requirements (partial revert of #2692).
- Use Python to parse whl METADATA.
- Bugfix the new simpler algorithm and add a new unit test.

Fixes #2830

(cherry picked from commit 5b9d545)

Summary:
- Better handle git references for sdists.
- Better handle direct whl references.
- Add an extra test that turned out to be not needed in the end, but I
  left it to increase the code coverage.

Work towards #2363
Fixes #2828

(cherry picked from commit a79bbfa)

It seems that the integration tests that I thought were covering this
had the same time. Added an assertion to the unit tests as well

Fixes #2863.

(cherry picked from commit 9dfa3ab)