Compare zipping performance of zipfile and zip

[tcompa]

Branching from #2215 to have a clean benchmark.

We created a heavy folder to be zipped:

$ find ./xxx/ -type f | wc -l
177818

$ du -sh ./xxx/
4.0G	./xxx/

And we zip this folder with two approaches.
In Python

from fractal_server.zip_tools import _create_zip
import time

t0 = time.perf_counter()
_create_zip(folder="/somewhere/xxx", output="/somewhere/zipbyfractal.zip")
t1 = time.perf_counter()
print(f"{t1-t0=}")

and with a native command

$ zip -qr zipbyzip.zip ./xxx/

Results

$ time zip -qr zipbyzip.zip ./xxx/

real	0m31.580s
user	0m24.308s
sys	0m2.715s

$ python zip_with_fractal.py 
t1-t0=45.33480614399923

$ ls -lh *zip
-rw-rw-r-- 1 tommaso tommaso 236M Jan 28 11:21 zipbyfractal.zip
-rw-rw-r-- 1 tommaso tommaso 246M Jan 28 11:20 zipbyzip.zip

TLDR:

• zip is indeed faster, on this specific case
• This difference is not enough to justify switching to zip right away, due to other costs (the actual implementation, and the additional external dependency which may not be available)
• Further explorations are possible, and especially by re-running the benchmark on a network filesystem (as opposed to the ext4 I used)
• There is a small difference in size of the two outputs, like due to the different compression level