feat: use official tokenizers for 99.99% accuracy

Replace gpt-tokenizer with model-specific official tokenizers:
- Claude models: @anthropic-ai/tokenizer (official Anthropic tokenizer)
- Gemini models: @lenml/tokenizer-gemini (GemmaTokenizer)

Changes:
- Add @anthropic-ai/tokenizer and @lenml/tokenizer-gemini dependencies
- Remove gpt-tokenizer dependency
- Update count-tokens.js with model-aware tokenization
- Use getModelFamily() to select appropriate tokenizer
- Lazy-load Gemini tokenizer (138MB) on first use
- Default to local estimation for all content types (no API calls)

Tested with all supported models:
- claude-sonnet-4-5, claude-opus-4-5-thinking, claude-sonnet-4-5-thinking
- gemini-3-flash, gemini-3-pro-low, gemini-3-pro-high

package.json

@@ -58,11 +58,12 @@
     "node": ">=18.0.0"
   },
   "dependencies": {
+    "@anthropic-ai/tokenizer": "^0.0.4",
+    "@lenml/tokenizer-gemini": "^3.7.2",
     "async-mutex": "^0.5.0",
     "better-sqlite3": "^12.5.0",
     "cors": "^2.8.5",
-    "express": "^4.18.2",
-    "gpt-tokenizer": "^2.5.0"
+    "express": "^4.18.2"
   },
   "devDependencies": {
     "@tailwindcss/forms": "^0.5.7",