[subset] TrueType/CFF2 variable font

[blueshade7]

• Add HVAR & VVAR subsetting for TT/CFF2
• Add gvar struct for TT (sanitize, look-up, subsetting)
• Calculate TT VF glyph widths (if HVAR/VVAR absent)
• Calculate TT VF glyph extents from glyf & gvar (no hint support)
• Add tests (subset: gvar, HVAR, VVAR)
• Add tests (TT VF widths & extents)
• ItemVariationStore subsetting in GPOS, GDEF
• Add tests (subset: VF GPOS, GDEF)

Not to subset: MVAR, BASE, cvar
Not in scope: axes subsetting, instantiation

[blueshade7]

Removed task "Calculate TT/CFF2 VF glyph widths from HVAR/VVAR" because it is already in place

[blueshade7]

This is my plan for HVAR & VVAR subsetting including VariationStore.

HVAR::subset to be called by _subset2; seems more convenient than _subset which requires to know size upfront (maybe rewrite CFF/CFF2 subsetter using _subset2 later?)

To subset inner indices for VariationStore, I am going to repurpose CFF::remap_t (one for each outer index), rewriting it based on hb_map_t & hb_set_t. Inner indices referenced by advMap/lsbMap/rsbMap are collected in hb_set_t, then old to new mapping is stored in hb_map_t.

I am not going to bother subsetting & remapping outer indices; for an unused outer index, its corresponding item variation data subtable is emptied. This wastes 6 bytes each but I don’t expect to see many outer indices in wild after all.

[blueshade7]

Actually an empty item variation data subtable wastes 10 bytes counting 4-byte offset. It would be nice if an item variation data offset may be set to null for a missing table, but OTF spec doesn't seem to say so. Darn.

[blueshade7]

Created branch var-subset for this task

[blueshade7]

Decided to subset outer indices & variation data subtables after all, as fontTools appears to support it.
Pushed HVAR & VVAR subset to var-subset branch.

[blueshade7]

Pushed advance width calculation using gvar & glyf

[blueshade7]

Pushed glyph extents calculation using gvar & glyf
Added api test test-ot-metrics-tt-var, subset full font test with a TT variable font

[behdad]

Closing since there's nothing actionable here.