Add `epi_archive` plotting (and animation?) functionality

[brookslogan]

autoplot should show revision history of a / multiple epikey-signal, potentially with some sort of paging functionality; Dan's made plots like these for various presentations, though that code may need generalization.

plot should probably forward to autoplot.

See also / use the accrued package for animations.

[brookslogan]

A basic implementation probably can be done in a day. Handling paging and/or adding to documentation probably makes this a multi-day task.

[dsweber2]

Definitely have wanted this several times (I ended up re-creating it in cmu-delphi/epipredict#431). I think the most awkward part is dealing with multiple keys; realistically, a plot with more than 3-4 sub-plots is ~impossible to read, and putting them on the same plot is also illegible. So I think any implementation is going to have to either expect the user to hand specific keys to plot, or choose either a random or representative subset e.g. one with the largest discrepancies, one with typical discrepancies, and one with the least discrepancies.

Related to #205 and revision_summary, could probably reuse some features for e.g. subplot choice.

[dajmcdon]

I was just searching for similar. Some thoughts:

1. The current .max_facets approach from epi_df "works" but is a bit annoying. Possibly add a selector for "which facets"?
2. What would you want the default to look like? Would this work as a first pass?

I would like 1 and 2 relatively ASAP, so I'm happy to put some cycles right now. Additional complexity can be deferred until later if still desired?

[dsweber2]

oh I was thinking of this plot from the docs overhaul, which I guess is like the 2nd plot from your slides? We might want to add some sort of flag for different kinds of plots. The "how to deal with different epi-keys" problem is true for either, and I think choosing based on worst/average/best case works for either.