The import report prints the same language tag twice when files disagree on casing

Importing PIC reports:

Languages: English, es-MX, es, es-MX

Not a double-count: those are two files whose gloss languages resolved onto one tag. The message
drops the only thing that explains it.

The converter groups by the raw GlossLanguage string, so es-MX and es-mx stay separate
groups; tag resolution then uppercases the region subtag, so both land on es-MX. The modal maps
each group to its resolved tag with no dedupe and no disambiguation
(Pt9ImportModal.tsx:91) — foldReport dedupes books one line above and not languages.

The report already carries merge.sameTagCollisions, which groups exactly these raw values.
Nothing surfaces it, so the explanation is computed and then discarded at the UI boundary.

Options: a footnote line when sameTagCollisions is non-empty, beside the existing
missing-books note — recommended, since the collision is a fact about the files rather than about
one language; or print the raw value beside a colliding tag (es-MX (es-mx)). Deduping the list
is cheapest and wrong: the groups really are separate and their glosses really did merge onto one
key.

Decide tagIsFallback at the same time — English prints identically to a real tag though the
report flags it — since the same line decides both.

Out of scope: whether two raw groups should share a gloss key. That's a converter question;
this is only about the report saying it happened.

Size: S. Priority: Low — cosmetic, but it reads as a double-count.

PIC is the reproduction (#283).

Start at Pt9ImportModal.tsx:91 and trace how foldReport maps language groups to resolved tags. Use merge.sameTagCollisions and the existing missing-books note as the entry point; done means the report explains non-empty collisions without deduplicating distinct groups, while deciding tagIsFallback in the same display logic.