Fix floating point issues

[farkmarnum]

This fixes #193 using the solution described in that issue.

I've also added a test for the case causing the error -- even though the error only occurs on some architectures, there's always a warning from numpy about the potential overflow, so the test checks for that.

[schlafly]

Thanks, that looks great!

[schlafly]

FYI, I need to diagnose the failing lon/lat tests ~now but they are unrelated to this PR.

[schlafly]

@farkmarnum , I resolved the issues with tests so now the tests work. Everything passes except for the -oldestdeps test, which fails.
https://github.com/spacetelescope/romanisim/actions/runs/12677867579/job/35334182610?pr=194#step:10:554
I don't immediately understand that.

[farkmarnum]

@farkmarnum , I resolved the issues with tests so now the tests work. Everything passes except for the -oldestdeps test, which fails. https://github.com/spacetelescope/romanisim/actions/runs/12677867579/job/35334182610?pr=194#step:10:554 I don't immediately understand that.

It looks like the behavior of np.nextafter changed somewhere between numpy v1.22.0 and the current version (v2.2.1), and a solution that works across both versions is to explicitly pass the dtype instead of just correctly typing the input:

I just pushed a fix to this branch.

[schlafly]

Great. Thanks! All tests are passing, merging.