PBR: derive diamond, calcite and glazed terracotta F0 from IOR, and zero diamond_ore emission #239

Non-blocking findings from the final review of #231 (LabPBR 1.3 material matrix), plus one inconsistency it left on the existing diamond_ore entry. Verified on main. The CHANGELOG.md:L19-L20 entry says gem, glass, quartz and obsidian F0 is derived from refractive index as ((n-1)/(n+1))^2 * 255, but several values do not follow that.

Gem F0

• Diamond (n = 2.417) gives F0 44. deepslate_diamond_ore uses 44, 48, 48, 50 and 40 across its five gem features (tools/lib/pbr-rules.json:L640, L648, L656, L664, L672), which work out to n ≈ 2.3-2.6. diamond_ore has the same five values (L67, L75, L83, L91, L99). Set every diamond gem feature on both entries to 44, the way emerald uses one IOR-derived value (13) for every gem feature.
• Update the assertions that pin the old values: tools/test/pbr-generator.test.mjs:L313-L316 (f0 >= 48) and L650 (Diamond crystal facet Green is F0 48). Assert iorToF0(2.417), like the emerald checks at L260 and L300.

Dielectrics with no stated derivation

• calcite uses F0 14 (tools/lib/pbr-rules.json:L889), which is n ≈ 1.61. Calcite averaged over nω/nε ≈ 1.57 gives 13.
• glazed_terracotta uses F0 14 (L917). A silicate glaze at n ≈ 1.5 gives 10-11.
• Derive both from IOR (or name the source for 14) and update tools/test/pbr-generator.test.mjs:L436 to use iorToF0. Add a glazed terracotta F0 assertion next to L465-L467.

Diamond ore emission

• diamond_ore gem features emit 20, 30, 30, 35 and 10 (tools/lib/pbr-rules.json:L69, L77, L85, L93, L101). Vanilla diamond ore emits no light. feat(pbr): expand LabPBR 1.3 material properties matrix (#53) #231 set deepslate_diamond_ore gem emission to 0 for that reason (asserted at tools/test/pbr-generator.test.mjs:L317-L320), and every other ore is 0. Set diamond_ore gem emission to 0 as well.
• Replace tools/test/pbr-generator.test.mjs:L652 (Diamond crystal facet Alpha has emission > 0) with an assertion that it is 0, and add the same every((f) => (f.emission ?? 0) === 0) check used for the deepslate variant. The synthetic emissive-core fixture in tools/test-shader-fixtures.mjs builds its own specular map and does not read pbr-rules.json, so it is unaffected.

Line endings

• .gitattributes has no rule for *.txt, so tools/test/fixtures/vanilla-1.21.6-block-stems.txt checks out with CRLF on Windows with core.autocrlf, against .editorconfig's end_of_line = lf. Nothing breaks today (the index copy is LF and the test splits on \r?\n). Add *.txt text eol=lf to the text block at .gitattributes:L2-L15.

Acceptance criteria

• Every gem, glass, quartz, obsidian, calcite and glazed terracotta F0 in pbr-rules.json equals iorToF0(n) for a stated n, and the tests assert it through iorToF0.
• No ore material emits light, diamond_ore included, and a test covers it.
• .gitattributes pins *.txt to LF.
• The CHANGELOG.md [Unreleased] entry matches the final values, and npm test passes.