diff3proof: add waterfall, glyphset and spacing/kerning proofs

[emmamarichal]

@simoncozens @m4rc1e

Summary

diff3proof currently only generates one proof type: paragraphs of sample text (--sample-mode context) or a codepoint-coverage pangram (--sample-mode cover). This PR adds three more proof types that existed in the old diffenator2/gftools pipeline but have no equivalent in diffenator3 today — including for the "onboarding" case (a single font, no "before" version to diff against):

• waterfall — the coverage pangram rendered at a ladder of point sizes (customizable via --waterfall-sizes), to spot hinting/ rasterization jumps.
• glyphs — every encoded glyph in the font, laid out in a grid.
• spacing — curated strings (ported from diffenator2's GF_Latin_Core Spacing/Kerning wordlists) that stress sidebearing consistency and kerning pairs. Latin-only for now; other scripts can follow the same pattern.

Design

--sample-mode is now repeatable (comma-separated), and all requested sections render into the same self-contained HTML file rather than one file per proof type:

diff3proof --sample-mode context,waterfall,glyphs,spacing font.ttf
diff3proof --sample-mode context,waterfall,glyphs,spacing before.ttf after.ttf

The default (--sample-mode omitted) is unchanged — same single context output as before.

Testing

• All modes, individually and combined, in both single-font and before/after mode
• --waterfall-sizes override
• Graceful degradation: a font with minimal Latin coverage correctly omits the Spacing & Kerning section entirely, rather than rendering empty boxes
• The existing before/after toggle (#fonttoggle) correctly flips all three new sections along with the existing ones, since they reuse the same .font-before/.font-after CSS classes
• Default behavior (no --sample-mode flag) verified byte-for-byte identical to main
• cargo fmt --all -- --check and cargo clippy --all-targets --all-features -- -D warnings both clean

Open questions

1. Single combined HTML file with sections (what this PR does) vs. one file per proof type + an index page, closer to diffenator2's layout.
2. Is Latin-only coverage acceptable for the spacing/kerning wordlist for now, with other scripts as follow-up work?
3. Should --sample-mode be renamed (e.g. --proofs) now that it no longer only selects a sample of text, or kept as-is for compatibility with existing gftools qa invocations?

Images