CT 2000 fix semver prerelease comparisons

[gshank]

resolves #6834
resolves #6461

Description

In semver.py we were using packaging.version.parse to parse the prerelease portion of versions, but since that version parsing is incompatible with SemVer, it started failing once the LegacyVersion fallback was removed.

This removes the dependency on packaging and subsitutes some comparison code specifically for prepreleases, so that we can remove the pin on packaging.

Checklist

• I have read the contributing guide and understand what's expected of me
• I have signed the CLA
• I have run this code in development and it appears to resolve the stated issue
• This PR includes tests, or tests are not required/relevant for this PR
• I have opened an issue to add/update docs, or docs changes are not required/relevant for this PR
• I have run changie new to create a changelog entry

[dbeatty10]

Highlights

• 🎉 This looks like it does as described and removes the dependency on packaging.version.parse()
• 💡 It seems like we should still swing back later and try to reduce our need for custom PEP 440 and SemVer parsing and comparison logic

More detail

I double-checked that CI indeed used packaging>=23.0:
https://github.com/dbt-labs/dbt-core/actions/runs/4068580210/jobs/7007304928#step:4:55

I gave a reading of the logic in compare() and _nat_cmp() and cmp_prerelease_tag(), and I didn't notice anything amiss or a key case not covered.

My only nits are the variable names prcmp (for "prerelease comparison"?) and _nat_cmp (for "natural comparison"?) felt a little terse and not 100% obvious. But given the time-sensitive nature of your fixes here, these didn't feel quite important enough to block or even raise as formal suggestions.

✅ Ultimately, it looks like the key test cases are covered by test__filter_installable.

Side note: I'm guessing black doesn't run against code in the test directory, and I won't raise any nits about white space formatting within test/unit/test_semver.py.

[gshank]

We should definitely look at a more consistent solution to these version problems, and look into using some existing package.

[jtcohen6]

Let's include this in the next v1.4.x patch release as well, since users will continue to have issues installing dbt-core with the tight pin. The change was small & precise, and fixed a narrow bug (the previous logic was genuinely incorrect).