v1.30.0

voxtype: override any config key for a single run

The headline change is --set KEY=VALUE on both voxtype and
voxtype post-process (#233). Until now, trying a different engine, model or
output format meant editing ~/.config/voxtype/config.toml, running once, and
editing it back. Each --set item now behaves exactly like that line in the
config file, for that run only:

voxtype post-process --set post-process-provider=cerebras \
                     --set post_process_format=minimal transcript.txt

Details worth knowing:

• Repeatable, and KEY accepts either hyphens or underscores.
• Values take the target field's type: true/false (also yes/no, on/off,
  1/0), numbers, strings, and lists as comma-separated items or a TOML array.
  A bare KEY= sets the empty value.
• Unknown keys and bad values exit 1 with a message and a "did you mean" hint,
  and every value still goes through Config.validate() — so a typo fails
  immediately instead of silently running with the old setting.
• Dedicated flags such as --engine take precedence over --set for the same
  key, so existing scripts behave as before.

Documented in --help, the voxtype-install skill, and the Starlight
configuration and overview pages.

Plugin packaging

• The voxtype Claude Code plugin is bumped to 0.2.0 so cached installs actually
  pick up the updated voxtype-install skill — the optional local/Cerebras
  transcript cleanup, and reading the API key from ~/.config/voxtype/.env
  (#232). Metadata only.

Verification

The voxtype suite passes (440 passed, 7 skipped), including 12 new tests in
tests/test_voxtype_set.py, plus a real post-process run against Cerebras.

Full changelog:
v1.29.7...v1.30.0