chore: pin plurimath to ~> 0.11.4

[ronaldtse]

Summary

• Tightens plurimath dependency from >= 0.10.1 to ~> 0.11.4 (latest release line)
• Adds Gemfile.lock to .gitignore (standard practice for gems)

Test plan

• bundle lock --update plurimath resolves cleanly against plurimath 0.11.4
• bundle exec rspec — 40 examples, 0 failures (95.25% coverage)
• CI green

[opoudjis]

Acknowledged.

Two notes on state first:

• main has been repaired to its last authorised state — plurimath ~> 0.10.1, version 1.11.1 (95427a4) — and the unauthorised tag v1.12.1 removed. This PR now reads as a clean ~> 0.10.1 → ~> 0.11.4 change from that baseline.
• The approach here is the right channel: a PR rather than a direct push, a bounded ~> 0.11.4 rather than an open-ended >=, with a stated test plan.

The remaining gate is rule 1 in #109: moving onto the plurimath 0.11.x line is exactly the class of change that must be tested against the entire metanorma stack, including the metanorma-cli samples — not html2doc's own specs alone. This will not be merged into main until I have run that full-stack testing myself; that testing, and the sign-off conditioned on it, remain my responsibility as maintainer.

As a courtesy, so the route to merge is unambiguous, the test is:

• a Gemfile.devel pinning html2doc to this PR branch (rt-update-plurimath);
• CI (GHA) runs against that Gemfile.devel for each downstream consumer — isodoc, metanorma-standoc, and metanorma-cli;
• metanorma-cli exercised across its full suite of sample repos (~3 hours) — the decisive gate.

No pass, no merge.

No objection to plurimath 0.11.x / MathML 4 as the direction — only to adopting it ahead of that testing.

🤖

[opoudjis]

Full-stack test result for plurimath ~> 0.11.4.

Tested via a Gemfile.devel pinning html2doc across isodoc, metanorma-standoc, and metanorma-cli. To isolate the plurimath change from an unrelated version-pin conflict, the pin points at a branch holding html2doc at the current 1.11.1 baseline with only plurimath ~> 0.11.4 (resolves to 0.11.5).

Results:

• metanorma-standoc — green across all matrix cells.
• isodoc — 5 failures / 378, consistent across every OS × Ruby cell, all in the number-format / MathML localisation subsystem:
  • numberformat_classification_spec — the new padding number-format option is not yet classified in isodoc's numberformat_type;
  • presentation_xml_numbers_override_spec (×3) — MathML number localisation output (large-notation min/max, data-metanorma-numberformat precision);
  • presentation_xml_numbers_spec — twitter_cldr_localiser_symbols fraction options.
• metanorma-cli — full sample suite still running (~3h); will follow up.

These are anticipated. The padding option was requested on the plurimath side; isodoc simply needs to classify it and align to the 0.11.x localisation output. I will address the isodoc side.

This is exactly why the change is gated on a full-stack run rather than the gem's own specs (rule 1, #109): a bounded but real set of isodoc regressions that html2doc's own suite does not surface. This PR merges once isodoc is updated for plurimath 0.11.x and the stack re-tests green.

🤖