I'm in favor of this! If someone is going to try this, I would also suggest doing some light reorganization of the test files so that it's clearer which ones go with which parts of the code.

For example, move a lot of the test files in the root directory that are associated with VASP IO into a vasp/ subdirectory, etc. Also I believe there are some old Q-Chem test files under qchem/ but the more relevant ones are under molecules/new_qchem_files. Perhaps these can be consolidated.

In PR #3146, I added a new OSZCIAR_MD to the test_files. @janosh you may want to include that into the compiled test files? But since now Oszicar can read fields dynamically in principle we can shrink two OSZICAR test files into one? I feel like there are many other duplicate test files can be shrunk into one too

I feel like there are many other duplicate test files can be shrunk into one too

I'm sure there are. Would be some effort to identify them but PRs welcome!

Can we close this now or there is more test files to be compressed or relocated that I could help with? @janosh

I guess it would be good to collect all cif (not sure about json) files together and I might do this in the near future 😃

What about the compress-test-files branch?

i think there are a lot of uncompressed output files left. try running compress_test_files.py from #3132 and it should show which ones i mean.

compress-test-files branch is super outdated but might help to keep it around for reference until we're all done