feat: add screenshot OCR URL detection using Tesseract.js

[shresthbhargava]

Closes #147

What's Changed

index.html

• Added screenshot upload section between the example chips and result div
• Drag & drop zone with visual feedback (border highlight on hover/drag)
• Hidden file input triggered by click or keyboard (Enter/Space) for accessibility
• Tesseract.js CDN loaded via jsDelivr (v5) — no npm install needed
• aria-live="polite" on result area so screen readers announce extracted URLs

script.js

Added full screenshot OCR URL detection module:

• handleDragOver / handleDragLeave / handleDrop — drag and drop handlers
• handleScreenshotUpload — file input change handler
• processScreenshot(file) — reads image via FileReader, draws to canvas, passes to Tesseract.js OCR with live progress %
• extractUrlsFromText(text) — regex extracts URLs from raw OCR text, deduplicates, strips trailing punctuation, filters noise (version numbers, too-short strings), caps at 10 results
• showScreenshotUrls(urls) — renders extracted URLs as a list with individual Scan buttons
• scanExtractedUrl(url) — fills the URL input with the extracted URL and triggers checkSecurity() automatically

style.css

• .screenshot-dropzone — dashed border drop zone with hover/dragover highlight
• .screenshot-divider — "or" divider between URL input and screenshot upload
• .screenshot-urls-found — card showing extracted URLs
• .screenshot-url-item — individual URL row with scan button
• .screenshot-scan-btn — styled scan button per URL
• .screenshot-processing / .screenshot-error / .screenshot-no-urls — state styles

Behavior

• User drags & drops or clicks to upload a screenshot (JPG, PNG, WEBP)
• Tesseract.js runs OCR entirely in the browser — no server calls, no data sent externally
• Progress percentage shown while OCR runs
• Extracted URLs listed with individual Scan buttons
• Clicking Scan fills the main input and triggers the existing security check flow
• If no URLs found, a helpful message is shown
• All interactions keyboard accessible

Note

Website Preview Sandbox (headless browser screenshot) is not included in this PR as it requires Puppeteer (~300MB server dependency). Can be implemented as a follow-up if the maintainer wants it.