Get rid of the shared memory arrays : add a colfile mmap option

[jonwright]

In sinograms/properties.py and sinograms/point_by_point.py the code works via shared memory.
This does not scale beyond one node and it fails on python2.7.

For global read-only memory we could use mmap with numpy on a non-compressed hdf5 file (https://gist.github.com/maartenbreddels/09e1da79577151e5f7fec660c209f06e):

assert dset.chunks is None and dset.compression is None and dset.is_virtual is None and dset.is_external is None and etc.
file = open(path, "rb")
fileno = file.fileno()
mapping = mmap.mmap(fileno, 0, access=mmap.ACCESS_READ)
np.frombuffer(mapping, dtype=dset.dtype, count=dset.size, offset=dset.id.get_offset()).reshape(dset.shape)

This may be useful for reducing some out-of-memory problems.

Another upgrade path could be looking into dask.dataframe for distributed processing.

[jonwright]

To make some progress try to break this up into a bunch of smaller tasks:

• Point by point code: write a hdf colfile. Each worker process reads it during pool init.

Properties.py is more challenging:

• Make lima_segmenter record the pixel -> peak labeling for both of the cp and lm schemes.
• Refactor lima_segmenter to write fewer files (e.g. one per process?).
• Labels will be saved with pixels.
• Peaks2d properties array (s1, sI, scI, srI, id) are available during segmenting to be saved with sparse pixels.
• Check the io speed / size with/without compression for saving pixels peaks. Pick something.
• Find and save the overlaps. This is one 'peaksearch' per overlap dimension. Output of a form (peak_i, peak_j, score).
• Determine the peaks3d labels across omega or dty
• Determine the peaks4d labels across the sinogram

[jonwright]

Note : multiprocessing + shared memory seems to be buggy. The remove from tracker monkeypatch does not work. Abandon it.

Exception ignored in: <Finalize object, dead>
Traceback (most recent call last):
  File "/cvmfs/hpc.esrf.fr/software/packages/linux/x86_64/jupyter-slurm/2023.10.7/envs/jupyter-slurm/lib/python3.11/multiprocessing/util.py", line 224, in __call__
    res = self._callback(*self._args, **self._kwargs)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/cvmfs/hpc.esrf.fr/software/packages/linux/x86_64/jupyter-slurm/2023.10.7/envs/jupyter-slurm/lib/python3.11/multiprocessing/synchronize.py", line 87, in _cleanup
    sem_unlink(name)
FileNotFoundError: [Errno 2] No such file or directory