BUG: fix float string literal issues

[bsipocz]

This logic was introduced in the big registry overhaul PR #289, and I see no review discussion for this part of the code.

Looking at it now, I think f-strings should be powerful enough when generating the SQL string. If not, we will definitely need some use case coverage in the tests, too.

First commit fixes #623, but I think the cleanup in the second commit should work, too.

The changes in numpy printoptions in the tests ensure that we indeed trigger the bug (it does have ample test coverage with existing tests). I expect some other places may be affected and will rely on CI to smoke out the version dependencies of them and will see whether they are bugs or the fixes can be included here, too.

[msdemlei]

I have not tried to fathom why unit tests didn't catch that one, and I can't find time to do that now. Similarly, I would have to spend more time than I have now to set up a numpy2 testbed, and so I have not actually ascertained that f-string interpolation will produce valid SQL literals. Since that seems very plausible, though, I will approve, but I am grateful if someone will do some stress testing of fix.

[bsipocz]

I have not tried to fathom why unit tests didn't catch that one,

We have opted in legacy printing in the test configs, so this all got effectively turned off for all unit tests, initially as a workaround, but apparently it had unintended consequences

I'm working on finding a fix that actually works for all the version and will push a commit for that, hopefully shortly.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 82.31%. Comparing base (c4080f7) to head (eba8c6f).
Report is 7 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #624      +/-   ##
==========================================
- Coverage   82.31%   82.31%   -0.01%     
==========================================
  Files          72       72              
  Lines        7431     7429       -2     
==========================================
- Hits         6117     6115       -2     
  Misses       1314     1314              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

---

🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[bsipocz]

OK, so with this latest commit I made a slight change, but I don't think it's making anything worse. Prior this we had the repr as "(23, 'Illuminatus')", and now we will make everything explicit string with "('23', 'Illuminatus')". Note that 23 was in fact already a string, and our other, default option would be to keep the numpy prints (as of "(np.int64(23), np.str_('Illuminatus'))", which I don't think is desirable here. cross-reference: https://numpy.org/devdocs/release/2.0.0-notes.html#representation-of-numpy-scalars-changed

Please re-review to say whether you are OK with this, or prefer if I find some other way to keep exactly what we had before.

[msdemlei]

With the caveats from my last review (plus I have not thought about whether DALQuery.__repr__ changes might break code): Let's go.

[bsipocz]

FWIW, I'm on np 2.0+ in all my dev envs (and this PR gets rid of the legacy print workaround, too), and we do have many of the notebooks/astroquery/etc CI, so those provide some actual testsbeds. I had a quickrun of their tests with this branch and it looked all OK, but I'll keep an eye on CI statuses for the longer running version.