Ensure alternative identifiers are added to flora.csv

[LukasWallrich]

alt_identifier_o | alt_identifier_r -- | --

[LukasWallrich]

Update — alternative identifiers are now wired into the pipeline (commit 08fafe3, on main).

How alt_identifier_o / alt_identifier_r are implemented:

1. Source columns (manual curation). They are optional columns on the source sheets — currently only present on the replications sheet — holding manually-curated alternative identifiers (e.g. a preprint DOI for a published work). The pipeline now selects them with any_of() rather than all_of(), so they are preserved when present and silently tolerated when a sheet doesn't have them.

2. Output. They are carried through to flora.csv in output_cols, positioned immediately after the corresponding DOI — doi_o, alt_identifier_o, … and doi_r, alt_identifier_r, …. (This replaces the old doi_o_alt / doi_r_alt names.)

3. Auto-populated by preprint dedup. When the preprint/publication dedup drops or replaces a row, the dropped DOI is appended to the surviving row's alt_identifier_* via append_alt_identifier_vec() — appended comma-separated and deduped case-insensitively, so multiple alternates accumulate instead of overwriting.

4. Gated on confirmation (important). Only confirmed override decisions (rows in cache/confirmed_preprint_duplicates.csv with action = keep_1 / keep_2 / keep_both) annotate alt_identifier_*. Auto-resolved pairs still drop/replace the duplicate row, but they do not write the alternate identifier until a human has confirmed the pair — we don't want to assert identifier-equivalence the reviewer hasn't actually checked.

5. Review nudge. To make sure unconfirmed auto-resolutions don't silently sit forever, maybe_open_dedup_review_issue() files (or comments on) a [preprint-dedup-review] GitHub issue when there are auto-resolved pairs and confirmed_preprint_duplicates.csv has gone stale (> 7 days).

Net effect: every alternate identifier in flora.csv is either manually curated on the source sheet or backed by a confirmed dedup decision. The columns will appear in flora.csv on the next pipeline regeneration.