Only assign memmap within boundaries for write_binary

[h-mayorquin]

This is like #1781 but for writing. I am writing the tutorial for how to use the job_kwargs and the memory heap reports crazy allocations for the following code:

from pathlib import Path
from spikeinterface.core import write_binary_recording
from spikeinterface.core import generate_recording



minutes = 30
recording = generate_recording(num_channels=384, durations=[60.0 * minutes], seed=0)


job_kwargs = dict(n_jobs=1, total_memory="1G", progress_bar=False, verbose=False)
file_path = Path(__file__).parent / "recording_test_n_jobs.dat"
write_binary_recording(recording=recording, file_paths=[file_path], **job_kwargs)

Reason being that ta memmap for the whole array (80 GiB) is reserved even if it not used. A possible consequence of this massive heap allocation is that maybe the system will overswap. With the current PR this heap allocations disappears.

This will make my life easier for tracking the efficiency of processing and I think is a small increase in complexity.

[h-mayorquin]

Performance:

this branch:

The main branch:

Changes are within the standard deviation of each other.

[samuelgarcia]

thanks. This is really cool.

[samuelgarcia]

Do we need this ?
the memmap_ojob is deleted no ?
This is a bit confusing because the file is stil open. Are we sure that this line do not close the underlying file ?

[h-mayorquin]

I am not sure but the file is not closed, otherwise the tests would not work.

[samuelgarcia]

Merci beaucoup beaucoup!
I ask for a small question.
We have lot of work to do on the neo side as well!!!