Parallel checks and debugging for empty filament cases

[e-koch]

• Additional debugging for longest path failures
• Check threading behaviour with 2D parallelization

Occasional edge cases continue to crop up in FIlament2D:

  File "/project//FilFinder/fil_finder/filament.py", line 274, in skeleton_analysis
    longest_path(edge_list, nodes,
  File "/project//FilFinder/fil_finder/length.py", line 446, in longest_path
    max_path_length = max(values)
                      ^^^^^^^^^^^
ValueError: max() arg is an empty sequence

• added debug=True mode for FilFinder2D.analyze_skeletons that enables additional printing to help identify these problem cases
• Always impose a minimum >1 pixel distance minimum. Otherwise the definition of a "filament" doesn't really make sense.
• Add a more clear error message for these cases.

[e-koch]

Changed the parallelization to adopt a single pool for processing, making it easier to change arguments for the pool and switch to using MPI:

e.g., examples added to the tutorial notebook:

from concurrent.futures import ProcessPoolExecutor

mypool = ProcessPoolExecutor(max_workers=2)
fil = FilFinder2D(hdu, save_name="FilFinder_Output", pool=mypool)

# is equivalent to

fil = FilFinder2D(hdu, save_name="FilFinder_Output", pool=None, nthreads=2)

# Using mpi4py:
# from mpi4py.futures import MPIPoolExecutor

# mypool = MPIPoolExecutor(max_workers=None)
# fil = FilFinder2D(hdu, save_name="FilFinder_Output", pool=mypool)

[e-koch]

Merging here. Tests and CI updates remain out-of-date and wil be fixed elsewhere.