Support GSUB type 6 format 2 with nested single substitutions

[mh0pe]

Description

Adds execution support for the GSUB chaining contextual substitution format that
the parser already exposes as lookup type 6, format 2.

The implementation:

• gates rules through coverage and the input class set;
• matches backtrack classes in reverse logical order, then input and lookahead
  classes through their separate class definitions;
• applies nested lookup records in design order, with later records consuming
  earlier substitutions;
• retains each sequenceIndex, so only the targeted token is changed;
• stops at the first matching rule and subtable, including deliberate no-op
  rules;
• supports nested single substitutions (formats 1 and 2), including type 7
  extension wrappers.

The nested-lookup scope is intentionally limited to single substitutions. Other
nested lookup types fail explicitly rather than being silently skipped. Lookup
flag filtering remains a broader shaping-engine follow-up; both real-world
reproducers below use lookup flag 0 and nested type 1.

Motivation and Context

On current master, the presence of a type 6 / format 2 subtable causes
getPath() and getAdvanceWidth() to throw before coverage is even evaluated:

substitutionType : 62 lookupType: 6 - substFormat: 2 is not yet supported

This reproduces with both SF Hebrew Rounded and the current Google Fonts Roboto
variable font.

This is a focused extraction of the GSUB 6/2 portion of #816, following the
review suggestion there to split the GSUB work from unrelated variation
changes. Thanks to @wiltse439 for the original implementation work; the commit
also credits them as a co-author. This PR leaves the existing type 6 / format 3
path untouched, preserving the sequenceIndex fix from #824.

How Has This Been Tested?

• npm test: 348 passing, including build, minified distributions, and source
  lint.
• Six synthetic regressions cover reverse backtrack order, first-match no-op
  rules, ordered records, nonzero sequenceIndex through token application,
  first matching nested and top-level subtables, extension wrappers, and
  explicit unsupported-type errors.
• SF Hebrew Rounded:
  • an end-of-run mark changes from glyph 110 to final-form glyph 163;
  • the same mark followed by a protected lookahead remains glyph 110;
  • the previous unrelated Latin probe renders without throwing.
• Google Fonts Roboto variable font:
  • i + combining acute shapes to glyphs [636, 169];
  • affinity renders without throwing.
• Both generated ESM and CommonJS distributions produce the same real-font
  results.

Screenshots (if appropriate):

Not applicable.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)

Checklist:

• I did npm run test and all tests passed green (including code styling checks).
• I have added tests to cover my changes.
• My change requires a change to the documentation.
• I have updated the README accordingly.
• I have read the Contribute README section.

[mh0pe]

Friendly maintainer ping @fdb @Connum — this is the focused GSUB type 6 format 2 extraction suggested during review of #816. It keeps the scope to nested single substitutions, fails explicitly for unsupported nested lookup types, and includes synthetic plus real-font coverage. I’d appreciate a review when time permits and am happy to adjust the scope.