Tests of fermi_hubbard in publication_test is not hermetic

[weinstein]

it downloads datasets from https://datadryad.org/stash/dataset/doi:10.5061/dryad.crjdfn32v
the download URLs are not super stable, but also datadryad.org website unavailability can cause the recirq test suites to flake out and fail

It may be better to avoid the external dependency on datadryad.org in tests and mock out the download.

[rmlarose]

Documenting discussion with the helpful folks at Dryad:

Our team has confirmed that individual files do not have their own unique link. Previously, the did but that was changed with the relaunch in 2019.

So a permanent fix should use the DOI of the dataset instead of URLs for individual files.