Try to fix test_skymax on Numpy 2

[titodalcanton]

This is an attempt at fixing #4990.

I am assuming compute_max_snr_over_sky_loc_stat() wants the input arrays to be in single precision for speed/memory considerations.

In my local tests, this will still fail because the results differ from the expected values by 1e-6 or less, so we may have to update the reference values.

[spxiwh]

This test is not worth having the test suite failing over, so I approve any patch that just gets this working again.

It's possible dynamic range stuff will make the simple fix not work unless dynamic range is applied.

It might also make more sense to just cast to complex64 at the 3 lines where we do load_frequencyseries

[titodalcanton]

Looks like I need to tell assertAlmostEqual() to be less strict. Are we content with 5 decimal places?

[titodalcanton]

The inspiral example is also failing:

>> [Mon Dec 16 14:10:41 UTC 2024] Compressing template bank
Traceback (most recent call last):
  File "/home/runner/work/pycbc/pycbc/.tox/py-search/lib/python3.11/site-packages/pycbc/scheme.py", line 221, in _scheming_function
    return _import_cache[mgr.state][func](*args, **kwds)
           ~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
KeyError: <function inline_linear_interp at 0x7f8b597c1580>

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/runner/work/pycbc/pycbc/.tox/py-search/bin/pycbc_compress_bank", line 191, in <module>
    hcompressed = compress.compress_waveform(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/pycbc/pycbc/.tox/py-search/lib/python3.11/site-packages/pycbc/waveform/compress.py", line 237, in compress_waveform
    hdecomp = fd_decompress(comp_amp, comp_phase, sample_points,
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/pycbc/pycbc/.tox/py-search/lib/python3.11/site-packages/pycbc/waveform/compress.py", line 441, in fd_decompress
    inline_linear_interp(amp, phase, sample_frequencies, out,
  File "/home/runner/work/pycbc/pycbc/.tox/py-search/lib/python3.11/site-packages/pycbc/scheme.py", line 238, in _scheming_function
    return schemed_fn(*args, **kwds)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/runner/work/pycbc/pycbc/.tox/py-search/lib/python3.11/site-packages/pycbc/waveform/decompress_cpu.py", line 38, in inline_linear_interp
    amp = numpy.array(amp, copy=False, dtype=rprec)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: Unable to avoid copy while creating an array as requested.
If using `np.array(obj, copy=False)` replace it with `np.asarray(obj)` to allow a copy when needed (no behavior change in NumPy 1.x).
For more details, see https://numpy.org/devdocs/numpy_2_0_migration_guide.html#adapting-to-changes-in-the-copy-keyword.

@GarethCabournDavies I think this is what you were describing on Slack? If you have a fix, could you push it here?

[spxiwh]

The decompress block of code is more important than the skymax test code ... We should check whether a copy should be being done there.

[spxiwh]

I applied Gareth's proposed fix here