Add `as.data.frame.epi_archive`, `as_tibble.epi_archive`, `as.data.table.epi_archive`, `reconstruct_archive` helpers

To avoid #618 while working around missing archive verbs more ergonomic, it will help to have these helpers:

Archive to data frame (or subclass):
- `as.data.frame.epi_archive`
- `as_tibble.epi_archive` (note: `%>% as.data.frame() %>% as_tibble()` due to `data.table` bug until version with fix is released & we require)
- `as.data.table.epi_archive` (note: do not alias input, or have argument to control)

Data frame (or subclass) back to archive:
- `reconstruct_archive`, similar to `dplyr_reconstruct`. (note: do not alias input, or have argument to control) (may just be able to destructure the `template` argument, then feed its metadata along with the `data` to `new_epi_archive`, `validate_epi_archive`?)

Start by reading the referenced epi_archive entry points, especially new_epi_archive and validate_epi_archive, and compare reconstruct_archive with dplyr_reconstruct. Check how each requested conversion helper should preserve archive metadata and avoid aliasing its input, including the noted data.table and as.data.frame() to as_tibble() constraints. Done means all five helpers exist with agreed conversion and reconstruction behavior, backed by tests for metadata preservation and input aliasing.

Written by the indexing model from the issue text.