Fix rayleigh not preserving input dtype

[djhoese]

This was discussed briefly on slack, but I noticed that the rayleigh correction in Satpy causes 32-bit float inputs to become 64-bit float outputs after correction. This is mostly due to two things:

1. The LUTs are 64-bit floats
2. The multilinear interpolator has a dtype argument that wasn't being specified and defaults to float64.

This PR fixes both and includes an update to the pre-commit config. I also added pyspectral to pre-commit.ci so the pre-commit config should be automatically updated for us.

NOTE: This PR doesn't make a huge difference in performance, but it doesn't hurt either. I think the consistency alone is worth it.

• Closes #xxxx
• Tests added
• Tests passed: Passes pytest pyspectral
• Passes flake8 pyspectral
• Fully documented

[codecov]

Codecov Report

Merging #195 (7630698) into main (e247cc2) will increase coverage by 0.03%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #195      +/-   ##
==========================================
+ Coverage   89.23%   89.27%   +0.03%     
==========================================
  Files          23       23              
  Lines        2536     2545       +9     
==========================================
+ Hits         2263     2272       +9     
  Misses        273      273              

Flag	Coverage Δ
unittests	89.27% <100.00%> (+0.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files Changed	Coverage Δ
pyspectral/rayleigh.py	98.36% <100.00%> (+0.03%)	⬆️
pyspectral/tests/test_rayleigh.py	100.00% <100.00%> (ø)

[mraspaud]

Pre commit is complaining, but otherwise lgtm