Fix header-ids extra generating duplicate ids (#661)

[apoorvdarshan]

Summary

The header-ids extra can emit duplicate id attributes (#661).

De-duplication relied only on a per-base-slug counter. When that counter is appended to a base slug, the resulting id can collide with a differently-named header whose own slug happens to equal the suffixed value:

# Chapter
# Chapter
# Chapter 2

produced:

<h1 id="chapter">Chapter</h1>
<h1 id="chapter-2">Chapter</h1>
<h1 id="chapter-2">Chapter 2</h1>   <!-- duplicate of the line above -->

The second # Chapter becomes chapter-2, and # Chapter 2 also slugifies to the base chapter-2, which was emitted verbatim.

Fix

header_id_from_text now tracks the set of ids it has already emitted. After building the candidate id it keeps bumping the base-slug counter until the id is genuinely unique, so the third header becomes chapter-2-2:

<h1 id="chapter">Chapter</h1>
<h1 id="chapter-2">Chapter</h1>
<h1 id="chapter-2-2">Chapter 2</h1>

The change preserves the existing -N suffix convention (e.g. duplicate Intro -> intro-2, as covered by header_ids_5) and the reset-count behavior for persistent Markdown objects; the new seen-set is reset in lockstep with the existing counter.

Tests / Verification

• Added a tm-cases regression test (header_ids_duplicate) using the exact repro from the issue. It fails on master (produced a duplicate chapter-2) and passes with this fix.
• Full test suite run: no new failures. The 14 remaining failures are all pre-existing and tagged knownfailure (unrelated to this change).
• Added a CHANGES.md entry under the unreleased section.

Disclosure: prepared with AI assistance; reviewed and verified locally.