refactor: complete v3 cleanup and critical bug fixes

Phase 1: Remove legacy v2 handlers and schemas
- Delete 6 legacy handlers (getMetadata, getPageStats, getToc, listImages, renderPage, readPdf)
- Delete 6 legacy schemas
- Remove pdfCacheStats from cache handler

Phase 2: Streamline to 6 core tools
- Register only: pdf_info, pdf_read, pdf_extract_image, pdf_ocr, pdf_search, _pdf_cache_clear
- Update server instructions and tool comments

Phase 3: Critical stability fixes
- Remove debug file I/O from render.ts (prevents disk filling)
- Add ReDoS protection to searchPdf.ts (100 char regex limit)
- Verify pdfDocument.destroy() cleanup in all handlers
- Remove useless try-catch in render.ts

Phase 4: Performance optimizations
- Replace inefficient regex character counting in pdfOcr.ts with loop
- Fix non-null assertion in character counting
- Verify tool naming consistency

Tests: Update integration tests and remove obsolete test files

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: Martin

dist/index.js

@@ -1,14 +1,6 @@
 import { text, tool, toolError } from '@sylphx/mcp-server-sdk';
-import { cacheClearArgsSchema, cacheStatsArgsSchema } from '../schemas/cache.js';
-import { clearCache, getCacheStats } from '../utils/cache.js';
-
-export const pdfCacheStats = tool()
-  .description('Inspect cache usage for text and OCR results.')
-  .input(cacheStatsArgsSchema)
-  .handler(async () => {
-    const stats = getCacheStats();
-    return [text(JSON.stringify({ stats }, null, 2))];
-  });
+import { cacheClearArgsSchema } from '../schemas/cache.js';
+import { clearCache } from '../utils/cache.js';
 
 export const pdfCacheClear = tool()
   .description('Clear text and/or OCR caches.')