Ucdn::$ot_languages has drifted from HarfBuzz's language tag table: 16 first tags differ and 800 languages are missing, e.g. ml, grc, qu

[jakejackson1]

Ucdn::$ot_languages (src/Ucdn.php:423, upstream the same table read by Otl::_getOTLLangTag() at src/Otl.php:6150) is a hand-kept subset of the language→language system mapping, and it has drifted from the table HarfBuzz generates (hb-ot-tag-table.hh: ot_languages2, ot_languages3, ot_languages3_multi, ot_languages3_blocked). #206 brings OtlTags::language()'s subtag handling in line with HarfBuzz 14.3.1, but deliberately leaves the table's existing entries alone, because syncing it changes the language system of existing documents.

I compared each of the 354 single-subtag keys the table had before #206 against HarfBuzz 14.3.1's tables, applying HarfBuzz's rule that a three-letter code it has no entry for is used upper-cased as its own tag:

	keys
same first tag	338
same first tag, but HarfBuzz lists more tags to fall back to	29 of those 338
first tag differs in HarfBuzz's table	4
no entry in HarfBuzz's table, and its ISO 639-3 fallback differs	12
in HarfBuzz's tables, not in Ucdn	35 two-letter, 772 three-letter

(After #206 adds the Chinese languages, zh, ga → IRI, IRT and nv → NAV, ATH, the last row is 33 and 753 and the second is 28.)

First tag differs (mPDF → HarfBuzz):

• ml MLR → MAL, MLR
• hy HYE → HYE0, HYE
• ber BER → BBR
• scs SLA → SCS, SLA, ATH

HarfBuzz's ISO 639-3 fallback differs: grc PGR → GRC, yid JII → YID, nso SOT → NSO, lua LUB → LUA, eot BTI → EOT, kvd KUI → KVD, mdc MLE → MDC, nco SIB → NCO, ril RIA → RIL, umb MBN → UMB, xom KMO → XOM, yso NIS → YSO.

More tags in HarfBuzz, which only change the result where a font lacks the first: e.g. mo → MOL, ROM; mnw → MON, MONT (hb-shape --language=mnw picks MONT from a font offering only that; mPDF uses the default); id → IND, MLY; ko → KOR, KOH; dv → DIV, DHV; iu → INU, INUK; tw → TWI, AKA; crj, crl, crm, cwd → their Cree tag, then YCR/LCR/TCR, then CRE.

Missing keys include two-letter ak, an, co, ia, in, iw, ji, qu, sc, sh, su, tl, za. These currently lay text out with the script's DFLT entry.

HarfBuzz also returns no tag for the 162 three-letter codes in ot_languages3_blocked and upper-cases any other code it does not list. Ucdn has neither rule. That matters for extended language subtags as well: after #206, one Ucdn has no key for keeps the language subtag's tags (ar-afb → ARA), which matches HarfBuzz for most of them.

Expected: generate the table from HarfBuzz's, the way utils/ucdn_update.php generates the Unicode tables, and follow HarfBuzz for these keys. The first-tag changes above alter existing documents, so the PR should list them.

[jakejackson1]

Two follow-ups from reviewing #213 that belong with the sync rather than that PR:

• After Select HarfBuzz's language system for a variant, script, region or extended language subtag, and for Chinese other than zh (#206) #213, Ucdn::$ot_languages values are a string for most keys and an array for ga and nv, and OtlTags::tags() casts. A generated table should make every value an array.
• Select HarfBuzz's language system for a variant, script, region or extended language subtag, and for Chinese other than zh (#206) #213 decides that a language is Chinese when its first tag is ZHS, ZHT or ZHH, and yue/lzh (ZHH/ZHT) take only the Hans rule. That matches HarfBuzz's generator today, but the sync should check it still holds, or name the Chinese languages explicitly.