fix(skill)!: rewrite text-effects bundle from scratch (license blocker)

Closes Rames's blocker on PR #991. The previous text-effects bundle
was vendored from `pixel-point/animate-text`, which the GitHub API
confirms has no license:

  GET /repos/pixel-point/animate-text         → license: null
  GET /repos/pixel-point/animate-text/license → 404

With no explicit license, default US copyright is "all rights
reserved" by the upstream author. Cannot ship under Apache 2.0.

This commit replaces the 48 vendored JSON files (8451 lines) with a
24-file rewrite (525 lines) using a fresh, simpler schema. The
effect IDs (typewriter, soft-blur-in, etc.) are concept names — not
copyrightable. The schema (`enter`/`exit`/`swap`, `target`,
`durationMs`/`staggerMs`/`easing`) is standard motion-design
vocabulary. Per-effect parameter values are picked from
motion-design intuition for each effect's intent — not copied.

### Structure changes

Before: dual directory + per-effect schema duplication
  assets/text-effects/effects/<id>.json   (~330 lines each, GSAP
                                           recipe + embedded spec)
  assets/text-effects/specs/<id>.json     (~40 lines each, portable
                                           contract — duplicates the
                                           spec section above)

After: flat single-file-per-effect with one shared rendering pattern
  assets/text-effects/<id>.json           (~20 lines each, just the
                                           per-effect parameters)
  references/text-effects.md              (shared GSAP rendering
                                           pattern, documented ONCE
                                           with split rules,
                                           CustomEase wiring,
                                           layout-aware exceptions)

The shared rendering pattern (split by target → set initial state →
stagger to enter.to → exit symmetric to enter → hard-kill at beat
boundary) is in text-effects.md instead of repeated in every effect
JSON. Sub-agents read the catalog once and the per-effect JSON gives
them the parameters. ~94% smaller file footprint, single canonical
implementation pattern.

### Per-effect schema

  { id, name, description, target, enter, exit, swap, notes? }

  target = "char" | "word" | "line" | "element"

  enter / exit = { durationMs, staggerMs?, staggerOrder?, easing,
                   from, to }

  swap = { mode: "crossfade" | "sequential", overlapMs, microDelayMs }

  notes = implementation guidance (when present)

  layoutAware = true for the 3 effects whose line container animates
                separately (kinetic-center-build, short-slide-right,
                short-slide-down)

### Reference updates

- references/text-effects.md: rewritten with our own catalog table,
  fresh duration / stagger values per effect, and the shared GSAP
  rendering pattern as one canonical section instead of per-effect
  recipes.
- step-3-storyboard.md: path references updated from
  `text-effects/effects/[id].json` → `text-effects/[id].json`.
  Removed the upstream-specific `showcase.library_adapters.gsap`
  reference; sub-agents now read the per-effect params + the shared
  rendering pattern in text-effects.md.

24 effects total, unchanged set: 7 per-character, 8 per-word,
2 per-line, 7 whole-element. The 3 layout-aware effects keep their
custom behavior documented in their JSON `notes` field and in
text-effects.md's layout-aware section.

Author: ukimsanov

skills/hyperframes/assets/text-effects/blur-out-up.json

@@ -0,0 +1,22 @@
+{
+  "id": "blur-out-up",
+  "name": "Blur Out Up",
+  "description": "Words arrive clean and exit upward with increasing blur. The entrance is matter-of-fact; the exit dissolves into atmosphere. Asymmetric pairing — good when the line should feel like it lingers in the viewer's memory rather than getting decisively dismissed.",
+  "target": "word",
+  "enter": {
+    "durationMs": 360,
+    "staggerMs": 90,
+    "easing": "cubic-bezier(0.22, 1, 0.36, 1)",
+    "from": { "opacity": 0, "y": 6 },
+    "to": { "opacity": 1, "y": 0 }
+  },
+  "exit": {
+    "durationMs": 520,
+    "staggerMs": 22,
+    "easing": "cubic-bezier(0.22, 1, 0.36, 1)",
+    "from": { "opacity": 1, "y": 0, "filter": "blur(0px)" },
+    "to": { "opacity": 0, "y": -28, "filter": "blur(12px)" }
+  },
+  "swap": { "mode": "crossfade", "overlapMs": 200, "microDelayMs": 0 },
+  "notes": "The 200ms overlap during swap is intentional — incoming text starts arriving while outgoing text is still mid-blur. Reads as a transition rather than a clean cut."
+}

skills/hyperframes/assets/text-effects/bottom-up-letters.json

@@ -0,0 +1,22 @@
+{
+  "id": "bottom-up-letters",
+  "name": "Bottom-Up Letters",
+  "description": "Letters rise from below in a pronounced staircase. Each character takes more visual time than per-character-rise — the motion is larger, slower, and reads as confident punctuation rather than ambient build.",
+  "target": "char",
+  "enter": {
+    "durationMs": 320,
+    "staggerMs": 65,
+    "easing": "cubic-bezier(0.18, 1, 0.32, 1)",
+    "from": { "opacity": 0, "y": 56 },
+    "to": { "opacity": 1, "y": 0 }
+  },
+  "exit": {
+    "durationMs": 280,
+    "staggerMs": 14,
+    "easing": "cubic-bezier(0.7, 0, 0.84, 0)",
+    "from": { "opacity": 1, "y": 0 },
+    "to": { "opacity": 0, "y": -22 }
+  },
+  "swap": { "mode": "crossfade", "overlapMs": 80, "microDelayMs": 0 },
+  "notes": "Pair with bold or display-weight headlines. The 65ms stagger creates an audible-feeling rhythm; works well when the headline lands on a beat marker in narration."
+}

skills/hyperframes/assets/text-effects/depth-parallax-words.json

@@ -0,0 +1,22 @@
+{
+  "id": "depth-parallax-words",
+  "name": "Depth Parallax Words",
+  "description": "Per-word entrance where each word enters at a different scale and slight vertical offset, simulating depth — back words start smaller and lower, front words larger and at baseline. Reads as a 3D-feeling layered headline without needing a real Z-axis transform.",
+  "target": "word",
+  "enter": {
+    "durationMs": 540,
+    "staggerMs": 110,
+    "easing": "cubic-bezier(0.22, 1, 0.36, 1)",
+    "from": { "opacity": 0, "y": 18, "scale": 0.82 },
+    "to": { "opacity": 1, "y": 0, "scale": 1 }
+  },
+  "exit": {
+    "durationMs": 360,
+    "staggerMs": 22,
+    "easing": "cubic-bezier(0.7, 0, 0.84, 0)",
+    "from": { "opacity": 1, "y": 0, "scale": 1 },
+    "to": { "opacity": 0, "y": -8, "scale": 0.92 }
+  },
+  "swap": { "mode": "crossfade", "overlapMs": 120, "microDelayMs": 0 },
+  "notes": "The scale + y combination is what reads as depth. Don't drop scale below 0.7 — at that point the word looks small rather than far. Keep above 0.8 for the parallax illusion to hold."
+}