Consider having separate `.ref_time_value` and `.version` in archive slides

[brookslogan]

(Part of #163.)

As part of an add-on to #477 designed to reduce some continued confusion/inaccuracies and column name clashes, I'm planning, in epix_slide(), to:

• rename ref_time_values to versions;
• produce a version output column rather than a time_value output column; and
• make both .ref_time_value and .version available to epix_slide() formula&tidy slide computations. Currently, they're the same thing.

They won't always be the same thing for archive slides/maps, though. E.g., if we

• relax the restriction that time_value and version belong to the same time_type, allowing, e.g., datetime versions
• add a generic archive -> archive slide function to get an archive with transformed signals (e.g. 7davs)
• allow version windows (for grabbing revision paths for visualization or modeling along the lines of this)

But when these differ, we should make sure that slide computations expressed as functions also can access both .ref_time_value and .version/.ref_version. But that may mean pushing/demanding epix_slide computation functions to accept this additional argument, potentially breaking a lot of downstream code and vignettes, etc., and/or causing the interface of the simpler cases and more complex cases to diverge; and this is just to make advanced stuff work rather than to fix difficulties with basic usage (unlike the planned changes above). So we should think/plan more carefully about this.