fix(convert): pass through <br> in markdown so table cells can have line breaks

[devlimits]

Summary

• Add br to the inline-HTML passthrough allow-list in MacroConverter, so <br> survives markdown→storage conversion instead of being escaped to <br>.
• <br>, <br/>, and <br /> all pass through; a <br> inside a code span/block stays escaped as literal text.

Why

confluence convert --input-format markdown --output-format storage (and the markdown input path of create / update) runs markdown-it with html: false, so all raw inline HTML is escaped. The converter works around this by stashing an allow-listed set of inline tags (u, sub, sup, mark, details, summary) around the render and restoring them afterward — but br was missing from that list.

As a result a table cell could not contain a line break. | v1 | 가<br>나<br>다 | produced <td><p>가&lt;br&gt;나&lt;br&gt;다</p></td>. This is the practical way to get a line break inside a table cell, since markdown table grammar allows only inline text per cell. With br added, the same input now yields <td><p>가<br>나<br>다</p></td>.

Behavior preservation

• False positives still rejected — the (?=[\s/>]) lookahead keeps <broken> and autolink-shaped <br:foo> out of the allow-list.
• Real escaping unchanged — only allow-listed tags are stashed; <, &, and non-listed tags (e.g. <script>) are still escaped.
• Code regions unaffected — <br> inside inline / fenced / indented code stays escaped as literal text via the existing code-range exclusion.

Test plan

• New tests in tests/macro-converter.test.js: <br> in a table cell passes through; <br/> / <br /> variants pass through; <br> inside inline code stays escaped.
• Full suite green (692 tests).

Out of scope

• The reverse direction (storage→markdown) already renders a <br> node as \n, so round-tripping a <br> back out of a table cell hits a separate, pre-existing markdown-grammar limitation. This PR only fixes the forward (authoring) direction.