SNOW-3570860 move to sa 20

[sfc-gh-mraba]

Please answer these questions before submitting your pull requests. Thanks!

1. What GitHub issue is this PR addressing? Make sure that there is an accompanying issue to your PR.

   Fixes #SNOW-3570860

2. Fill out the following pre-review checklist:

   • I am adding a new automated test(s) to verify correctness of my new code
   • I am adding new logging messages
   • I am adding new credentials
   • I am adding a new dependency

3. Please describe how your code solves the related issue.

• Drop SQLAlchemy 1.4 compatibility code and require SQLAlchemy 2.0+.
• Update supported Python versions to 3.9 through 3.14.
• Remove obsolete SQLAlchemy 1.4 test paths, CI matrix entries, and compatibility helpers.
• Keep UUID reflection available on SQLAlchemy 2.0 and return UUID values as strings.
• Clean up SQLAlchemy 2-only reflection and ORM compiler paths.
• Move unreleased changelog entries under the v2.0.0 section.

[sfc-gh-jcieslak]

Q (for potential future PRs): Do we have other flags that could be removed in favor of better defaults?

[sfc-gh-mraba]

This is open for discussions before RC release.

[sfc-gh-jcieslak]

Q: Just so I understand correctly, we're skipping those as we have our own tests that work with the dialect-level flag, so we don't need the tests provided by SQA for truediv, correct?

[sfc-gh-mraba]

Good catch — the skip reason is incorrect. SA compiles // to FLOOR(x / y), not a SQL // operator, so "Snowflake does not provide //" is wrong. Snowflake supports FLOOR() fine.

The real issue is force_div_is_floordiv=True (the current default):

• truediv / tests expect 15/10 = 1.5 — these pass because visit_truediv_binary delegates to SA's base which emits 15 / CAST(10 AS NUMERIC) = 1.5
• floordiv // tests expect 15//10 = 1 — these would also pass via FLOOR(15 / 10) = 1, just with a PendingDeprecationWarning

So the skips are likely unnecessary. However, this is tied to the open force_div_is_floordiv discussion — if we flip its default to False in v2.0.0 (which I'm leaning towards given we're doing a major version bump), the warning disappears and the entire TrueDivTest subclass can be dropped. Leaving this pending until Q2 is resolved.