Add structured PDF and Office understanding

[interlinkglobal]

Priority 3 — structured document understanding

Implements roadmap item #67 as a read-only extension of Interly's existing sensitive read_text_file path.

Supported formats

• PDF: pages, text, heuristic headings, tables, metadata
• Word .docx: ordered headings, paragraphs, tables, core metadata
• Excel .xlsx / .xlsm: sheets, used ranges, headers, rows, formulas
• PowerPoint .pptx: slides, titles, text blocks, tables, speaker notes, core metadata

Safety and privacy

• preserves existing Y = terminal only / A = explicitly allow Groq access semantics by reusing read_text_file
• read-only source handling
• 50 MB document input ceiling
• Office ZIP entry and uncompressed-size checks before parsing
• bounded page, slide, sheet, table, and total extracted-text limits
• corrupt/encrypted Office packages fail closed

Packaging

• adds pinned document backends to package metadata and requirements
• explicitly bundles PDF/Word/Excel/PowerPoint backends with PyInstaller
• adds --check-document-support for frozen-build verification
• Windows workflow smoke-tests document backend imports from the built interly.exe

Verified

• 104/104 Windows tests passed, including generated PDF, Word, Excel and PowerPoint fixtures
• Ruff passed
• standalone interly.exe built successfully
• frozen executable passed --version and --check-document-support
• Inno Setup installer built successfully
• WinGet manifests generated successfully
• Windows distribution artifact uploaded successfully
• roadmap updated to 89/101 and #67 marked complete

Priority 3 remains active with #66 Recycle Bin deletion and #70 malware scanning/quarantine still open.