fix(linter): end index link paths at their own closing paren

[fohara]

Closes #60.

Problem

extract_markdown_link_path in the orphan rule took a link destination from the first ]( to rfind(')') — the last parenthesis on the line rather than the one closing the link. An index entry with a trailing parenthetical:

- [Alpha](tasks/alpha.md) (historical, superseded)

yielded the path tasks/alpha.md) (historical, superseded, which matches no file, so alpha.md was reported as W_INDEX_ORPHAN despite being right there in the index.

Two links on one line broke both: [Alpha](a.md) and [Beta](b.md) yields a.md) and [Beta](b.md, whose last component still ends in .md, so the .md guard passed and a nonsense path was recorded as a legitimate reference. Using find for the opening delimiter also meant only the first link on any line was considered.

Fix

• display::extract_link_paths scans a line for every link, ending each destination at the parenthesis that closes its link, with nested parentheses balanced and angle-bracketed destinations unwrapped. display::extract_link_path (already doing the forward scan for a single link) now delegates to it.
• The orphan rule collects all destinations on a line and keeps those naming a Markdown file or a directory. A CommonMark link title (path.md "Title") is stripped, but only as a fallback — a bare path containing spaces is tried whole first, so [A](tasks/my file.md) still resolves.

Tests

• Unit tests on both scanners: trailing parenthetical, mid-sentence parenthetical, immediately-adjacent parenthetical, two links on one line, nested parens, titles, angle brackets, spaces.
• End-to-end lash lint tests over the issue's exact project: no W_INDEX_ORPHAN, while a genuinely unreferenced file is still flagged exactly once.