feat: Reuse component IDs and preserve explicit XML line breaks

[u-00a0]

XML-to-JSON conversion currently replaces internal component IDs on every render, turning small edits into document-wide diffs. Add optional referenceJson rendering options that match existing widgets, tabs, documents, blocks, list items, table axes/cells, audio items and video elements before allocating new IDs. Exact content and unique local identities take priority; compatible edits retain IDs, while ambiguous additions receive fresh IDs. XML stays free of identity attributes.

Also preserve text newlines as closed <br></br> elements and parse explicit breaks back into inline newlines. Previously an empty <br> was ignored, concatenating archive text. The containing paragraph and its ID now remain intact. This prepares version 0.6.0; no package publication is part of this PR.

Validation:

• pnpm check: typecheck, 64 tests and build pass.
• Actual item 2338: all 24 widgets, 52 documents and 915 blocks retain IDs through conversion.
• Laevat source → generated XML → Wiki JSON: all 14 basic-record newlines match the source, 136 closed breaks survive, and all 915 block IDs remain unchanged.
• Regression coverage includes insertion/deletion/reordering, cross-container moves, duplicate content, local namespaces, table merges, collision retries, deep nesting, formatting and escaped markup.

The consuming extension was verified with 80 unit tests and a mocked Chromium E2E flow, including actual audio/quick-edit submission bodies. Full captured page data remains local; only a minimal public archive excerpt is included as a regression fixture.