un-whitelist deprecation warnings

[hamogu]

We have whitelisted a number of deprecation warnings and others in our testing. I believe that some of them are not longer relevant and thus should be removed. I'm pushing this as a PR so we get CI to run and see if any of our current tests fail with this.

Some of them date back to the Python 2.7 era according to a git blame.

Why do I care? I just had tests fail locally with an "np.asscalar(a) is deprecated" and I wondered why. I looked into Sherpa and did not find the string "asscalar" anywhere (because @DougBurke probably fixed that years ago). So, why did my test fail locally? Well, I had a recent version of numpy where assaclar has actually been removed and an older version of one of our other dependencies (astropy), which still used it. Since we whitelist this warning (even after Sherpa itself has been fixed) I;d never noticed that that particular environment still had such an old version of astropy.

So, I might as well try and see how many of those warning we can remove from our whitelist and still pass our current matrix of tests.

[codecov]

Codecov Report

Merging #1877 (f2adb28) into main (b230164) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #1877   +/-   ##
=======================================
  Coverage   81.24%   81.24%           
=======================================
  Files          73       73           
  Lines       26461    26461           
  Branches     3962     3962           
=======================================
  Hits        21497    21497           
  Misses       4803     4803           
  Partials      161      161           

Files Changed	Coverage Δ
sherpa/conftest.py	93.05% <ø> (ø)

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b230164...f2adb28. Read the comment docs.

[DougBurke]

Ideally I'd like the warnings explicitly hidden at the call site, as that would

a) not hide the warnings from other code
b) let us know exactly where they happen and make changes more obvious

but it's a lot of work to do (and I'm seeing some bizarre recwarn behavior that may depend on pytest version and almost-certainly on my understanding of things), so I don't see this happening any time soon.

So just removing the warnings is a subtle way to implement this!

There are two use cases that we can't check here

• a user with a random set of modules who is trying to test their build of Sherpa
• the CIAO test builds of Sherpa

but you and I are exemplars of the former, and I'm hoping that our CI runs are close-enough to the CIAO build systems that it shouldn't be a problem., If there are differences then it suggests that funky things are afoot that we should look at anyway....

[hamogu]

Since CI passed, I'm hoping that this is good. I've asked for a CM review (@Marie-Terrell was suggested by GH first in the list before I saw @harlantc but input fomr either of you is welcome).

I suggest to put this in sooner rather than later, so that we have a little time left to reactivate any of those white listings if we see warnings and failing tests cropping up in some obscure (i.e. not CI tested) combination of version dependencies.

[DougBurke]

This whitelist approach is always going to be something we have to review from time to time,so thanks for doing it. I'm a lot happier nowadays as we have

a) a reasonable coverage of build setups on CI
b) more tests than we used to

[Marie-Terrell]

This looks good to me. I'll double check this in CIAO once it goes in

[hamogu]

And to @DougBurke I made this exactly because I had a random assortment of packages (in this case numpy 1.25 and astropy 5.0) and in numpy 1.25 the function in question was actually removed because the deprecation period was over, but my astropy was so old that it still called np.asscalar. There is no way to make that right except by upgrading astropy, so if a user (e.g. me) has an environment with sherpa 4.16 and dependencies that are so old that these warnings are thrown (i.e. at least four numpy verison back), I think we want the warning to be shown to entice users to upgrade whatever dependency throws that warning.

The other whitelistings I removed where put in 7 years ago in the Python 2.7 era and we know Sherpa does not run on 2.7 any longer anyway.