Add direct image input support

[amirgholipour]

What changed

• Accept PNG, JPEG/JPG, TIFF/TIF, and PDF uploads in the current React and Express UI.
• Validate file signatures as well as extensions before uploading to S3.
• Preserve the detected content type and distinguish direct images from PDFs.
• Send direct images to classification as one-page inputs without PDF conversion.
• Update the agent workflow and UI documentation for the supported formats.

Why this change is proposed

The upload path currently assumes every input is a PDF. Image-only documents therefore require an external conversion step even though the downstream specialists already operate on images. Relying on the browser-provided MIME type alone would also allow an extension/content mismatch.

This change makes the upload boundary understand the formats the processing pipeline can already consume. Signature checks keep validation server-side, and direct images skip an unnecessary conversion stage.

Benefits

• Supports common scanning and archival formats directly.
• Reduces preprocessing and latency for single-page images.
• Avoids quality loss caused by wrapping an image in a PDF before analysis.
• Preserves the existing PDF path and current UI architecture.

Validation

• Tested PNG upload and direct-image processing against a deployed BADGERS environment.
• Verified PDF behavior remains available.
• Verified production UI build with pnpm.
• Reviewed the branch against the current upstream main.

Authored by Saeed Kasmani (amirgholipour).