support zlib compression

[magland]

This adds support for zlib compression when creating datasets, so you can do something like:

ds0 = g.create_dataset(
    f"spike_counts_ds_{ds_factor}",
    data=spike_counts_ds.astype(np.int32),
    chunks=(np.minimum(num_bins_per_chunk, num_ds_bins), num_units),
    compression='zlib'
)

This is important because zlib compression (unlike gzip and blosc) is deterministic (given the compression level). So then, if zlib compression is used, then two lindi.tar files that were created in the same way are going to be byte-equivalent. This is relevant for when I have a script that generates a file and stores in a content-addressable database (like kachery).

[rly]

Note to self: Python gzip is not deterministic, as I had thought, because it adds a timestamp

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 11.11111% with 8 lines in your changes missing coverage. Please review.

Project coverage is 81.46%. Comparing base (65491bf) to head (90725f8).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
...indi/LindiH5pyFile/writers/LindiH5pyGroupWriter.py	0.00%	7 Missing ⚠️
lindi/conversion/h5_filters_to_codecs.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #101      +/-   ##
==========================================
- Coverage   81.68%   81.46%   -0.23%     
==========================================
  Files          30       30              
  Lines        2774     2784      +10     
==========================================
+ Hits         2266     2268       +2     
- Misses        508      516       +8     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.