t27f: Cyrillic idea factory -- UTF-8 ideas set in a 5x7 particle font #10788

Part of epic #10560. Follows the idea factory (#10771, PR #10784).

The idea factory reads its idea as ASCII and sets it in scene_raster's 3x5 font, which has no Cyrillic. The owner's reels are Russian. Today the Russian reels (lesson_reel, token_floor) draw their text with a TTF font through ffmpeg drawtext. That text is not made of particles and cannot move to the FPGA path. A 3x5 cell is also too small for Cyrillic letters such as Zh, Sh, Shch and Yu.

Boundary

• New: specs/t27f/glyph57.t27.
  • UTF-8 decoding: cp_len, cp_at and cp_count. A malformed byte reads as one byte.
  • Case folding: Latin a..z, Cyrillic a..ya and yo fold to upper case.
  • The 5x7 font:
    • glyphs for ASCII 32..95 and for the 21 Cyrillic capitals with no Latin look-alike;
    • the 11 look-alikes (A, Ve, Ie, Ka, Em, En, O, Er, Es, Te, Ha) and Yo reuse Latin glyphs through a 32-entry map;
    • an unknown code point draws as ?.
  • Integer only: lookups and compares.
• Changed: specs/t27f/idea_factory.t27.
  • It wraps by characters, not bytes.
  • It sets text in the 5x7 font, with a 6-pixel advance and a line pitch of 9.
  • It accepts UTF-8 ideas, written with \xNN escapes because L3 keeps sources ASCII-only.
• Out of scope: lowercase glyphs, voice, and Verilog synthesis.

Scenarios

• An idea in Russian, for example "particles build the frame" written in UTF-8, wraps by letters. A 7-letter Russian word, 14 bytes long, stays on one line.
• Lower- and upper-case Russian light the same cells.
• The English sample idea still gives 7 lines, 2 scenes and 216 frames.

Requirements

• Decoding is pinned by tests: Zh (U+0416) is D0 96; yo (U+0451) folds to Yo (U+0401); a stray continuation byte counts as one character.
• No two stored glyphs share a bitmap, and the space glyph is blank.
• Every look-alike letter lights exactly the cells of its Latin twin.
• Every target of every scene stays inside the frame. A scene lights at most MAXCELLS cells.

Criteria

• t27c test-report passes on both specs with 0 vacuous tests (Railway lab).
• Seals are saved and verified.
• tools/dupe_scan.py is clean.
• A Russian preview reel is rendered on the lab and shown to the owner.

🤖 Generated with Claude Code

Requirements

FR-001: The glyph57 spec MUST provide UTF-8 decoding with cp_len, cp_at, and cp_count where a malformed byte reads as one byte.
FR-002: The glyph57 spec MUST implement case folding for Latin a..z, Cyrillic a..ya, and yo to upper case.
FR-003: The glyph57 spec MUST define a 5x7 font with glyphs for ASCII 32..95 and the 21 Cyrillic capitals with no Latin look-alike.
FR-004: The glyph57 spec MUST map the 11 look-alike Cyrillic capitals (A, Ve, Ie, Ka, Em, En, O, Er, Es, Te, Ha) and Yo to their Latin glyphs through a 32-entry map.
FR-005: An unknown code point in the glyph57 spec MUST draw as ?.
FR-006: All lookups and comparisons in the glyph57 spec MUST use integer arithmetic only.
FR-007: The idea_factory spec MUST wrap text by Unicode characters, not bytes.
FR-008: The idea_factory spec MUST set text in the 5x7 font with a 6-pixel advance and a line pitch of 9.
FR-009: The idea_factory spec MUST accept UTF-8 ideas written with \xNN escapes.
FR-010: Decoding MUST be pinned by tests: Zh (U+0416) is D0 96; yo (U+0451) folds to Yo (U+0401); a stray continuation byte counts as one character.
FR-011: No two stored glyphs in the glyph57 spec MUST share a bitmap, and the space glyph MUST be blank.
FR-012: Every look-alike letter in the glyph57 spec MUST light exactly the cells of its Latin twin.
FR-013: Every target of every scene in the idea_factory spec MUST stay inside the frame, and a scene MUST light at most MAXCELLS cells.

Success Criteria

• t27c test-report specs/t27f/glyph57.t27 passes with 0 vacuous tests and exit 0
• t27c test-report specs/t27f/idea_factory.t27 passes with 0 vacuous tests and exit 0
• Seals for specs/t27f/glyph57.t27 and specs/t27f/idea_factory.t27 are saved and verified with exit 0
• tools/dupe_scan.py runs clean with exit 0