Force the compression method of R feather saving 'uncompressed'

[alex-l-kong]

What is the purpose of this PR?

There have been issues with consistency in saving the .feather files in Python vs R. To prevent interoperability issues, we change the compression argument passed into write_feather in R to 'uncompressed'

[ngreenwald]

It wasn't the compression. Getting around the same number of errors, now with a different error message.
. Given that I used the MIBIAN a fair amount beforehand without running into these errors, it seems likely that its either the multiprocessing update or the switch to 3.7 that is causing these issues

[alex-l-kong]

@ngreenwald weird, I'll have to look into this some more. I wonder if it's an OS-related issue with MIBIAN since neither myself nor Candace has run into this issue locally. Just to clarify, on MIBIAN, have you tried this process on other datasets and found the same or similar errors pop up?

[ngreenwald]

This never popped up on my test dataset. I'm going to rerun clustering there to confirm that the issue comes up now. If so, it means it was one of the recent changes. Can you work with @srivarra to get a python 3.6 image (tagged with something besides latest) pushed to dockerhub so that I can rollback and see if that resolves it?

[alex-l-kong]

@ngreenwald I'm going to spend some time on the MIBIAN this afternoon to test it out on Candace's dataset and compare with the old and new Docker images when they both are posted. Will let you know if anything comes up.

[ngreenwald]

This was resolved in a previous PR