Refactor blocked/parallel reprojection

[astrofrog]

This refactors the handling of blocked/parallel reprojection with dask following a conversation with @Cadair.

• Ensure that we clean up all temporary directories with Numpy memmaps and zarr arrays
• Add a return_type option to reproject_interp which can be 'numpy' (default) or 'dask'. When the latter is used, nothing is computed when reproject_interp is called so this is very fast
• Document that to use schedulers other than synchronous or processes, users should set return_type to 'dask' and then compute it with the desired scheduler (e.g. a dask distributed scheduler). However, we need to make it very clear that the code is not thread-safe, so that multi-threaded schedulers can't be used (I don't think there is a way for us to check that)
• Expand the blocked/parallel reprojection to the adaptive and spherical intersect algorithms
• Make sure we properly test roundtrip_coords and order - these were not getting passed through to the blocked/parallel reprojection so were getting ignored, but need to add tests to make sure these options do something

[codecov]

Codecov Report

Merging #374 (1b8b7c0) into main (1432cec) will decrease coverage by 0.49%.
The diff coverage is 93.33%.

@@            Coverage Diff             @@
##             main     #374      +/-   ##
==========================================
- Coverage   93.86%   93.37%   -0.49%     
==========================================
  Files          24       25       +1     
  Lines         847      861      +14     
==========================================
+ Hits          795      804       +9     
- Misses         52       57       +5     

Impacted Files	Coverage Δ
reproject/interpolation/core.py	96.61% <ø> (+1.08%)	⬆️
reproject/adaptive/core.py	95.91% <83.33%> (-1.86%)	⬇️
reproject/common.py	91.08% <91.08%> (ø)
reproject/adaptive/high_level.py	100.00% <100.00%> (ø)
reproject/interpolation/high_level.py	100.00% <100.00%> (ø)
reproject/spherical_intersect/core.py	98.61% <100.00%> (+0.53%)	⬆️
reproject/spherical_intersect/high_level.py	100.00% <100.00%> (ø)
reproject/utils.py	83.33% <100.00%> (-3.68%)	⬇️

... and 1 file with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[Cadair]

As mentioned in chat we should have a way of parallelising over independent celestial input chunks as well.

[astrofrog]

@Cadair - I am working on that but will add it in a separate PR. In the mean time, would you be able to review this or are you happy for me to just merge?