SaveAsDeduplicated method

Rewrites the document as a single clean revision that first sweeps unreachable objects and then collapses compatible fonts, images and form XObjects onto one shared representative

PDFium first writes a non-incremental copy of the current in-memory document, then the wrapper performs a document-wide reachability walk and drops every unreachable object exactly like SaveAsReachable

Fonts, images and form XObjects are then fingerprinted: each candidate dictionary, its full reachable subgraph and its raw stream payload are hashed recursively, so duplicates merge only when their entire content is identical

Every cross-reference that pointed at a merged-away duplicate is rewritten to the surviving representative, and object numbers and generations stay stable so every surviving cross-reference remains valid

The output is a single classic revision with no object streams and no cross-reference stream, and the rebuilt trailer carries /Root, /Info and /ID forward

• TPdfDedupReport reports kept and pruned object counts plus per-category merged font, image and form XObject counts, with source and output byte sizes and a structured failure message
• Encrypted documents are rejected because a single-revision rewrite cannot honour the object numbers, byte ranges, and per-object keys that encryption depends on
• Signed documents are rejected unless AllowSignedDocument is explicitly enabled because a full rewrite invalidates existing signature byte ranges
• Candidates are grouped per resource type so an 8-byte hash collision can never collapse a font onto an image or form

SaveAsReachable, SaveAs, SaveAsNormalizedContent