v0.65.3

Extract text from remote PDFs and binary documents.

• Remote PDFs now extracted via pypdf instead of being stored as decoded binary garbage
• Remote DOCX and PPTX also extracted via temp file → FileDocumentProvider
• Other binary types (images, audio, video) get a descriptive placeholder
• Temp files written to ~/.cache/keep/ to satisfy path-traversal guard