fix(release): keep a changeset's markdown in the changelog

build_section piped every changeset body through `tr '\n' ' '`. A multi-line
changeset came out as one paragraph, so nested bullets rendered as inline
" - " runs and a whole release read as a single unbroken block — v0.3.0 was one
~4 KB bullet.

The renderer is now Ludic rather than a shell one-liner. A section is grouped
by conventional-commit type (Features, Fixes, Performance, ...), each changeset
is one bullet, and continuation lines are indented two spaces so nested lists
and paragraphs stay inside their item. Bullets are sorted within a group, so
cutting the same release twice produces the same text.

Also:

- `x release --dry-run` renders the pending section to stdout and touches
  nothing, so a release can be read before it is cut.
- `x changelog-render` re-renders a section from a directory of changesets, and
  `x changelog-section` prints one release's section back out of CHANGELOG.md.
- The v0.1.0 and v0.3.0 sections are re-rendered with the former, from the
  changesets recovered at each tag's parent commit; the bullet counts (6 and
  25) and word multisets are unchanged. v0.2.0 is left alone: it carries a
  hand-written summary and topical subheadings, and regenerating it would have
  replaced curation with raw changeset dumps. The file header now says that
  a section may carry such a summary, since it previously claimed released
  sections are never hand-edited.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

changes/README.md

@@ -16,10 +16,36 @@ the changelog. Markdown is fine.
 - `bump:` — `major`, `minor`, or `patch` (SemVer). The release version is bumped
   by the **highest** level among the pending changesets (unless `x release <level>`
   overrides it).
-- `type:` — the Conventional Commit type (`feat`, `fix`, `perf`, `docs`, …); it
-  becomes the bold prefix of the changelog bullet.
+- `type:` — the Conventional Commit type (`feat`, `fix`, `perf`, `docs`, …). It
+  decides which group the change lands in: `feat` → **Features**, `fix` →
+  **Fixes**, `perf` → **Performance**, and so on, in that order. A type with no
+  known heading gets one named after itself.
+
+## Writing the body
+
+The body is markdown and reaches the changelog as markdown: it becomes one list
+item, with continuation lines indented to stay inside it. Nested bullets, blank
+lines between paragraphs and inline code all survive.
+
+```
+bump: minor
+type: feat
+**Tiled map support** — load and draw Tiled maps.
+
+- **TMX/TSX** — the XML formats, decoded to the same intermediate as JSON.
+- **Collision** — the `collision` layer projects onto the engine tilemap.
+```
+
+Lead with the thing that changed, not with the mechanism. A reader scanning the
+release should be able to stop after your first clause.
 
 ## Adding one
 
 Create a file with a short, unique name, e.g. `changes/regex-namespace.md`. Any
 filename works except this `README.md`, which the release step always skips.
+
+Preview how the next release will read before cutting it — this writes nothing:
+
+```bash
+x release --dry-run
+```

changes/changelog-structure.md

@@ -0,0 +1,14 @@
+bump: minor
+type: fix
+`x release` keeps a changeset's markdown intact. Bodies used to go through
+`tr '\n' ' '`, which flattened every multi-line changeset into one paragraph —
+nested bullets came out as inline `" - "` runs and a release read as a single
+unbroken wall of text. A section is now grouped by change type (**Features**,
+**Fixes**, **Performance**, …) with one bullet per changeset and continuation
+lines indented to stay inside it.
+
+- `x release --dry-run` renders the next section to stdout and writes nothing,
+  so a release can be read before it is cut.
+- `x changelog-section <version>` prints one release's section from
+  `CHANGELOG.md`; `x changelog-render` re-renders a section from a directory of
+  changesets. The v0.1.0 and v0.3.0 sections were re-rendered with these.