ENH: use Usher for Pangolin assignment

[AnonymousUserUse]

It seems that issues with the server used for training the model of pangoLEARN (https://github.com/cov-lineages/pangolin-data/releases/tag/v1.19) cannot be solved in a short time. Besides, an assignment with Usher is generally more accurate than that with pangoLEARN. Many public health institutions, including UKHSA, RKI, and China CDC, have decided to use Usher in favour of pangoLEARN.
Therefore, I suggest either using Usher for the Pangolin assignment in the CoV-Spectrum website to replace the pangoLEARN assignment, or extending a further assignment possibility in the CoV-Spectrum website by Usher assignment.

[chaoran-chen]

Thank you very much for the suggestion and sorry that I am responding very slowly at the moment! It is definitely a good suggestion to include UShER. However, we are currently rewriting most parts of our backend and I'm afraid that this task will have to wait until then. (Very unfortunately, we will have to calculate the UShER assignments ourselves for the GISAID data and cannot just take pre-computed values which would make the implementation much easier.)