TreeInfo can use alternative Pango classifications

[hyanwong]

To match Ana's new lineage imputation metadata keys.

Relevant to #134 as it raises an error if the metadata key isn't present.

Instead of saving both designations in the TreeInfo class, you have to pick one at the start. I thought that would reduce the amount of output from the various tabulating functions, and I think we probably want to pick just one classification or the other, not use both. But happy for this to change, obviously.

[hyanwong]

Note that not all samples have a GISAID pango classification, so I have put "unknown" for these (I thought I might as well use lowercase, to distinguish from the imputed "Unknown").

[hyanwong]

Good point. Will change.

[hyanwong]

Perhaps just "pango_source" would do?

[jeromekelleher]

Sure

[hyanwong]

Done