Draw the p, f, k and kh ligatures with l and m, and release 3.103 - #50

What

The ligatures of p, f, k and kh with l and m: sixteen drawings (_PL, _PM, _FL, _FM, _KL, _KM, _K2L, _K2M, each .init and .medi), the twenty-four ligature glyphs the OTL composer builds from them, a rebuilt features.fea, and the version bumped to 3.103.

and the same in medial position. The word-final l and m keep the nirugu.extend insertion, as they do now: no ligature is drawn for them, following BL/BM.

These are a choice of this font rather than something the standard asks for — GB/T 25914—2023 does not list them — so the drawings are in the unified source and the ligatures are taken by the unified composer alone, not by the ligature table the standard fonts share.

Follows up on:

• 3.100: vertical origin moved to the ascender, upright punctuation misplaced in vertical text #49 — the reply about p / f / k / kh before l and m: the drawings were welcome if someone made them, and the OTL would follow.
• Draw the ligatures of p, f, k and kh with l and m Kushim-Jiang/mongolian#59 — the drawings, and the choice above.

Also in the rebuilt features.fea

The marked form of o / u / oe / ue that an initial consonant asks for is now carried across a consonant cluster to the medial vowel only, as GB/T 25914—2023 tables B.13/B.14 have it. A word-final vowel after a word-initial consonant cluster keeps its default form (n n oe → u1825.U.fina, not u1825.Ue.fina), while a single initial consonant still marks it (n oe → u1825.Ue.fina). This is the fix for Kushim-Jiang/mongolian#60.

Verification

• The source compiles, and each of the eight pairs ligates in initial and medial position.

• Shaping the Mongolian suites, exactly four cases change — the ligature replacing the nirugu.extend insertion, and nothing else:

  Case	Before	After
  core-hud > syllabic-o/u-5	u183A.K2.init nirugu.extend u182F.L.medi …	u183A_u182F.K2L.init …
  core-mag > pu2-bowed	u183A.K2.init nirugu.extend u182F.L.medi …	u183A_u182F.K2L.init …
  eac-hud > MPS11-51	u182B.P.init nirugu.extend u182F.L.medi …	u182B_u182F.PL.init …
  eac-hud > XKS11-52	u183B.K.init nirugu.extend u182F.L.medi …	u183B_u182F.KL.init …

  The stored expectations in qa/shaping_tests/ are updated for those four; every other case is unchanged.

• Version bumped to 3.103 (both versionMinor and openTypeNameVersion).

Note

The rebuilt features.fea comes from the mongfontbuilder working tree, which carries the o/u/oe/ue fix beyond the pinned 0.15.3. requirements.txt still pins 0.15.3, so the pin needs the release that carries the fix before make shaping reproduces these locks. As before, scripts/gen_shaping_tests.py is not in the tree; the locks here were regenerated by replaying the shaping suites directly.