Make epix_slide more like reframe

[brookslogan]

Resolves #275, addresses _pkgdown.yml typo, adds GitHub Actions checks to dev branch PRs, implements current approach to #213, adds missing test-grouped_epi_archive.R file.

[brookslogan]

• The last point in Change epix_slide to be more group_modify-like #275 still needs to be checked. [punted this decision and always output ungrouped tibble for now.]
• NEWS.md needs to be updated.
• Fix DESCRIPTION version to be 0.6.0.9999
• Resolve check failures
• Fix not getting epi_dfs out of some non-all-versions (at least ungrouped ones; check also on grouping by geo) [punted]
• Fix documentation/implementation in light of reframe always outputting an ungrouped object.
• Clean up branch issues from accidentally PR-ing off of fork
• Debug as_list_col = TRUE actually outputting an edf/tibble col bundle

[brookslogan]

Sometimes we can get an epi_df out of epix_slide as I believe was originally intended. But not very elegantly:

• We don't output epi_df when ungrouped
• We output epi_df when grouped by geo_value, but if slide computation was trying to output an epi_df or add a key column, it's ignored/corrupted by default, or raises errors with names_sep=NULL.

And if we want an epi_df out of epix_slide with time_values from the slide computations, not the ref_time_values, we may be stuck trying to wrap epi_dfs in our computation inside a list column of a tibble output, then attempt to unnest and deal with naming conflicts simultaneously, which may not be possible, so we may have to rename & lose any outer epi_df-ness, then unnest, then add an other_keys entry for whatever the ref_time_values label was. Or, alternatively, we don't wrap, do the rename shuffling, then have to reclass manually and reset metadata, some of which would have been discarded in intermediate steps (which means hardcoding or breaking the pipe to assign to a temporary to allow later access; both awkward).

I'm not going to try to make this easier, because it again raises questions of whether we really do want an epi_df out sometimes, or if we want something else. Perhaps we should only output a tibble like with all_versions=TRUE, or consider alternatives such as another epi_archive or a generalized archive. And revisit #163.

[brookslogan]

This PR is based off of a branch from the wrong remote. It is replaced by #311.