Warn if a non-`version` key is duplicated in `epi_df` or `epi_archive`

[nmdefries]

Currently, we do not check for distinct key values. This has some potential uses but also introduces the opportunity to form unexpected or invalid archives. Continue allowing this behavior, but warn the user if duplicate keys exist when creating or adding to an archive or an epi_df.

[brookslogan]

#89 discusses this as well. This issue is definitely the first step to do. However, we may later decide to either (a) turn it into an error, or (b) make things actually work when there is a warning. Actually, since this breaks the key as_of function, and thus also epix_slide, as well as the current epix_merge defaults (+ makes the user have to do post-processing to make non-defaults work) and stops the overhauled epix_merge in #101 from running, I think maybe we do want to just go ahead and make it an error. Probably not helpful to construct an epi_archive with this condition if most things don't work and something works but requires manual hacking to fix up issues.

[brookslogan]

If we do keep it a warning, we need it to be a dire warning like "Current epi_archive methods were built assuming the key was a unique key; e.g., epix_as_of and epix_slide will give incorrect results, and other methods may also malfunction."

[brookslogan]

epi_archive side of this is handled in lcb/grouped_epi_archive. Just made it a hard error since key methods like as_of malfunction otherwise. Decision might differ for epi_df.

[brookslogan]

Being addressed in #519

[nmdefries]

Addressed in #519