[PERF] Make new_epi_archive more unsafe and more efficient and use it in safe places

[dshemetov]

Alright, I added a validate_epi_archive and moved a bulk of the validation from new_epi_archive in there. Now as_epi_archive calls validate and then new. new_epi_archive is only used there, so this change should be fine for now, but in the future we can refactor some safe internal calls to as_epi_archive with new_epi_archive. There is still a bunch of validation and construction of the data.table object in new_epi_archive, so that's another thing that could be fixed.

I also made clobberable_versions_start default to NULL and then I set it to NA in that case. Just went with the consistent option for now, though we can easily switch it later.

Originally posted by @dshemetov in #431 (comment)

[brookslogan]

As part of making compactify_tol more accessible [to more conveniently test some other changes], I'm rearranging things to better align with the pattern here and cleaning up some duplicate validation code. I don't see it being used within other package code except for epix_merge(); I won't touch that for now. [So I doubt we'll get performance benefit from this refactor.] (Seems clobberable_versions_start's default is NA now, so no worries about when NULL -> NA should be handled.)