support: core agent should read pdfs from any channel

[AggManik]

Problem: CORE agent currently can’t read/act on PDF attachments (e.g., emailed PDFs, Slack uploads, WhatsApp docs, etc.).

Task: add first-class PDF support so the CORE agent can ingest a PDF provided in any channel, extract text (and optionally structure like headings/tables), and then run normal agent reasoning/actions on it.

Requirements:

• Works across channels (email attachments, Slack files, WhatsApp documents, web uploads).
• PDF parsing pipeline (OCR fallback for scanned PDFs).
• Preserve page numbers + citation offsets so agent can reference source.
• Enforce size/time limits and safe handling (malicious PDFs).
• UX: agent should acknowledge receipt and ask clarifying question only if extraction is low confidence.

Acceptance notes:

• MVP: text extraction for text-based PDFs.
• V2: OCR for image/scanned PDFs + table extraction.

Labels: enhancement.