Enable compression for hash_table

[ArvidJB]

As a follow up to #205, can we enable compression for the hash_table datasets?

Ideally this should be configurable in some way, maybe as an argument to VersionedHDF5File(f, hash_table_compression='lzf')? If this turns out to be difficult it's also okay to use some default compression for all hash_table datasets.

[ArvidJB]

Also, the hash_table chunk sizes should be configurable. I see that Hashtable does take a chunk_size argument, but unfortunately that's not exposed to the user.

[asmeurer]

The hash table looks the same regardless of what is in the dataset. So I would think it's better to just find a compression that works and use it everywhere.

[ArvidJB]

Both 'gzip' and 'lzf' seem to work and are part of the h5py install:
https://docs.h5py.org/en/stable/high/dataset.html#lossless-compression-filters
We usually pick 'lzf' for everything other than dtype='object'. Maybe we can just go with 'lzf' as the default?

[quarl]

lzf sounds good, let's go with that for now.

Aaron:
We are trying to meet a tight deadline and need a release with this fix by tomorrow.
Is it better for you to make the change or Arvid?

[asmeurer]

We are trying to meet a tight deadline and need a release with this fix by tomorrow.

Sorry, didn't see this comment until just now. I have a fix at #211. I will merge it an make a release as soon as the tests pass.

[asmeurer]

I didn't make the chunk size configurable yet. Hopefully that isn't also something you need done urgently. I'm actually not even sure if the hashtable dataset needs to be chunked at all. I might need to play with this.

[ArvidJB]

Thanks, just the compression helps us a lot already!
Thanks for the quick work and also making a release already!