Backend Model Update for the FINAL FINAL time

[danich1]

The resource used for our word2vec models (Pubtator Central) underwent a major update. This update amplified some of the earlier years (e.g., ~10k to 579k available documents) allowing them to be used again. This also means the backend models need to be updated once again to reflect this change. Plus, we are going to drop the abstracts as that is redundant to fulltext. I'm currently rerunning them now to make sure everything is complete.

In terms of todo items this is all that arises:

• Drop all abstract models (will appear in next PR for word-lapse-models by me)
• Upload the last set of models (will appear in next PR for word-lapse-models by me)
• Re-generate the cache again with all full text documents
• Prepare to have preprint models available (will upload as soon as I can)