[hardening] PdfParser silently produces empty output for scanned / image-only PDFs

[MTCMarkFranco]

Enhancement

PdfParser.ParseAsync produces an empty ParsedDocument when given a scanned / image-only PDF (no text layer). Downstream the pipeline silently reports zero findings, which is indistinguishable from a clean document — a serious failure mode for real customer uploads.

Acceptance criteria

• After parse, if total canonical text length is 0 AND the PDF has ≥1 page with non-zero image content, throw a typed ScannedPdfNotSupportedException (or equivalent) with a clear message.
• Alternatively (and tracked here as one acceptable resolution): emit the failure into ParsedDocument.Metadata with key parser_warning and add an integration check that surfaces it in CLI / API output.
• Decision documented in the implementation PR.
• Unit test with a small image-only PDF fixture.
• Engine genericity guard still passes.

Out of scope

• OCR integration (would be its own feature, not hardening).