refactor: retain bundle metadata for documents

[AlexanderLanin]

Why

Sphinx discovers document names independently of the bundle metadata that describes their source. We need that document-to-bundle association available in Python so later processing can relate document-level data to the corresponding bundle. PR #841 provides the metadata; this stacked PR carries the association through the documentation build without consuming it yet.

What

• Preserve the bundle-only manifest for standalone Needs actions.
• Map mounted document names to their bundle metadata and associate otherwise-discovered root documents with the root bundle.
• Keep pure data mounts out of the document mapping.
• Expose the mapping to later consumers and cover the behavior with focused tests and documentation.

[MaximilianSoerenPollak]

Closed on purpose?