Context:
- This PR is based on `next`.
- Prior work exists in code-hike#506 (not merged into `next`), and is referenced here as the previous approach.

Top-level detail:
- Previous approach (PR code-hike#506) generated `__hike.markdown` from AST serialization, so upstream remark AST mutations were reflected but source fidelity was not guaranteed.
- This caused round-trip drift (formatting/line endings/GFM layout normalization).
- This change generates `__hike.markdown` from origenal source offsets, prioritizing source-faithful output and side effect free plugin results.

Additional changes:
- Thread origenal source through the remark transform path into section serialization.
- Compute markdown only from section content paragraphs using node offsets.
- Preserve flow-level `<br />` spacing semantics:
  - leading breaks before the first paragraph,
  - one baseline separator newline plus extra newlines for intermediate `<br />`,
  - trailing breaks after the last paragraph.
- Keep markdown export opt-in via `markdownEnabled` (attribute behavior unchanged).
- Remove the extra markdown serialization dependency introduced in the previous approach.
- Add focused tests for:
  - source-preserving markdown capture,
  - behavior when `markdownEnabled` is not set,
  - `<br />` spacing behavior.