Iterate on compactify_tol interface & usage

[brookslogan]

Currently, it's hard to set compactify_tol since we have to go through the less-convenient new_epi_archive. We should provide compactify_tol as an arg of as_epi_archive.

The default compactify_tol may not be appropriate if the user has per-capita rate data (rather than e.g. per-100k); it may be safer to use 0 --- and make 0 actually work; dplyr::near checks < tol and we'd need <= tol. This also prevents surprises when performing epix_slides that should preserve compactness and converting to epi_archive still talking about compactifying.

Really, compactify and compactify_tol shouldn't even be in new_epi_archive to begin with according to tidy design or some other prominent guidelines, since new_* are supposed to do just simple class validation checks.

[brookslogan]

Also, we probably shouldn't be using tol when looking at the epikeytime columns. E.g., if we are using daily counts and decide a difference of a few counts can be neglected, we don't want a tol, say, in the 1--3 range to cause us to potentially skip time_values altogether in flat periods.