Folio upload/picker can't handle folios with no chant-start rows in Cantus DB (e.g. CH-Fco Ms. 2, folio 003r)

Some manuscript folios have no chant text starting on them — all their text belongs to the previous folio's chant, since Cantus Database only records where a chant starts, not every page it continues onto. Such folios never appear in Cantus Database's folio list for that source, and since landing-page's folio picker is built entirely from that list, these real, imaged folios currently can't be processed at all: batch uploads get blocked by a "counts must match" error with no way past it, and single-image uploads simply can't select the folio from the dropdown.

Real example: CH-Fco Ms. 2, folio 003r (Cantus source 123672) — a real page, invisible to Cantus Database's list because no chant begins there. First reported against mothra-text (DDMAL/mothra-text#42), but the fix belongs here: mothra-text already handles this folio correctly once it gets the image in the right position — it's landing-page's upload flow that can't get it there.

**Proposed solution:** for batch uploads in manuscript order, landing-page will auto-detect a "hidden" folio like this when it falls between two known folios, slot it in automatically, and badge it (e.g. "not in Cantus source — continuation") so it's visibly flagged. For single-image uploads, and batches where the folio's neighbors aren't included, the folio picker will gain a "custom folio…" option to type the label by hand.

**Limits:** only folios with an actual uploaded image and a clean single-step gap get recognized — it can't run the pipeline on a folio with no image. It doesn't verify the gap is truly a continuation (vs. a blank/non-chant page); those would still get processed, with quality issues surfacing like any other alignment problem.