fix(core): address white-box review feedback

- Fix unicode surrogate pairs and emoji slicing via Array.from in ORP calculator
- Add slot smoothing constraint to pivot element to prevent horizontal jitter
- Implement adaptive timestamp compensation and visibilitychange auto-pause
- Slim storage layer by updating only STORE_META on progress changes with memory fallback
- Align EPUB chapter indices to token array indices instead of raw character offsets
- Support standalone emoji tokens via Extended_Pictographic regex
- Code-split JSZip via dynamic import to optimize initial bundle size

Co-Authored-By: Claude Code <noreply@anthropic.com>

src/core/orp.ts

@@ -1,6 +1,8 @@
 /**
  * Optimal Recognition Point (ORP) Calculation for Spritz RSVP Reading.
  * Fixates the eye at ~30%-35% into the word where human ocular fixation naturally rests.
+ * Uses unicode-safe code point segmentation (Array.from) to eliminate surrogate pair
+ * and emoji splitting defects.
  */
 
 const CJK_REGEX = /[一-龥぀-ヿ가-힯]/;
@@ -15,9 +17,11 @@ export function isCJKWord(word: string): boolean {
 
 /**
  * Calculates the 0-based index of the ORP character in cleanWord.
+ * Based on unicode code-point length (Array.from(cleanWord).length).
  */
 export function calculateOrpIndex(cleanWord: string): number {
-  const len = cleanWord.length;
+  const chars = Array.from(cleanWord);
+  const len = chars.length;
   if (len <= 1) return 0;
 
   if (isCJKWord(cleanWord)) {
@@ -45,6 +49,7 @@ export interface SplitWord {
 /**
  * Splits a token into prefix, pivot (ORP character), and suffix.
  * Handles leading punctuation (e.g. "“", "《", "(") and trailing punctuation.
+ * Operates purely on unicode code points to prevent corrupting surrogate pairs.
  */
 export function splitOrp(
   coreWord: string,
@@ -60,10 +65,11 @@ export function splitOrp(
     };
   }
 
+  const chars = Array.from(coreWord);
   const orpIndex = calculateOrpIndex(coreWord);
-  const prefix = leadingPunct + coreWord.slice(0, orpIndex);
-  const pivot = coreWord[orpIndex] || '';
-  const suffix = coreWord.slice(orpIndex + 1) + trailingPunct;
+  const prefix = leadingPunct + chars.slice(0, orpIndex).join('');
+  const pivot = chars[orpIndex] || '';
+  const suffix = chars.slice(orpIndex + 1).join('') + trailingPunct;
 
   return {
     prefix,