Fixed wheel choice by interpreter version tag

[mentalisttraceur]

Fixes #32 .

There seem to be unrelated test failures already in the automated test suite, but those failures are the same both before and after my patch, and the tests I added (and the one test I changed to factor out some now-reused code) are all passing.

[mentalisttraceur]

Local testing was a struggle. I couldn't find any documentation on it, but between the various errors and tracebacks and the definition of the GitHub PR testing action, I figured out that I needed to

1. create a dist directory in the repo root,
2. download and unpack a Pyodide release into it (so it should look like ./dist/pyodide/...), then
3. create a conftest.py with a DIST_PATH which is a pathlib object that points to that dist folder, and finally
4. run pytest with --dist-dir=./dist/.

That got most of the tests running locally.

Eight of the tests I couldn't get running (and five of those are also broken in the github action, and are failing in PR #33 as well). They seem orthogonal to this patch and fail the same way both before and after my patch.

[mentalisttraceur]

Since tests/test_micropip.py wasn't already using hypothesis, and since it would take much longer, I resisted the temptation to pull it in and write tests with Hypothesis strategies for generating creative variations of wheel versions and tags.

Maybe that's a PR for a later time!

[hoodmane]

Hypothesis strategies for generating creative variations of wheel versions and tags.

Sounds great!

[rth]

Thanks a lot @mentalisttraceur for the fix and for writing tests!

Yes, we should definitely add contributing docs.

A couple of minor comments otherwise LGTM. Could you please merge upstream/main in; the CI issues were fixed in #35

[mentalisttraceur]

Cheers! Requested changes made and tests passing. Let me know if I missed anything else. Thanks for the guidance, info, and taking care of your side of things so quickly!

[hoodmane]

Thanks @mentalisttraceur!