feat: scrub PDF metadata after redaction

[federicodeponte]

Summary

• Adds a scrubMetadata() helper in frontend/src/engine/pdf.ts that clears all standard PDF info dictionary fields (Title, Author, Subject, Keywords, Creator, Producer, CreationDate, ModDate) using MuPDF's Document.setMetaData() API
• Removes XMP metadata by deleting the Metadata stream from the PDF catalog via PDFDocument.getTrailer() low-level access
• Calls scrubMetadata() in both applyRedactions and applyPseudonymisation right before saveToBuffer, ensuring all redacted output is metadata-clean

Closes #4

Implementation details

MuPDF's WASM TypeScript bindings expose:

• Document.setMetaData(key, value) for clearing info dictionary entries
• PDFDocument.getTrailer() and PDFObject.delete() for removing the XMP metadata stream from the catalog

Both operations are wrapped in try/catch for graceful degradation if specific keys or low-level operations are unsupported in a given MuPDF build.

Test plan

• Upload a PDF with metadata (Title, Author, etc.) and run redaction; verify output has no metadata via pdfinfo or a PDF metadata viewer
• Upload a PDF with XMP metadata; verify the XMP stream is removed after redaction
• Run pseudonymisation on a PDF with metadata; verify output is metadata-clean
• TypeScript type checking passes: tsc --noEmit (verified locally)