v2.0.0 — Sharding + verdict-based polling

New: sharding. Add shard-count (2–40, grep/affected modes) to run tests in parallel across shards and merge into one Checksum run. See the README "Sharding" section.

New output: verdict. The exit code now gates on the server-computed verdict (pass/fail) rather than the raw status string — correct for sharded runs, empty selections, and infra failures.

⚠️ Behavior changes (why this is a major bump):

• An empty test selection (grep matched nothing) now fails the step instead of passing.
• wait: true now polls the run-status endpoint by run id instead of by job name.
• test-run-id is now set at dispatch (previously only after a terminal status).
• A 404 from the status endpoint now fails fast instead of retrying.

Pin @v1 for the old behavior; move to @v2 to adopt the above.

v1.0.0

First public release of the Checksum AI test-run action.

Quick start

```yaml

• uses: checksum-ai/test-run-action@v1
  with:
  api-key: ${{ secrets.CHECKSUM_API_KEY }}
  grep: 'checkout'
  auto-heal: true
  ```

What it does

Wraps the Checksum AI public-API execution endpoints in a single GitHub Action.
Supports four execution modes (`grep`, `suite-ids`, `test-ids`, `collection-id`),
optional auto-heal-on-failure, and optional opt-in polling that mirrors the
test-run terminal status into the workflow's exit code.

Highlights

• Auto-resolves `repo-name` from `github.repository` and `pr-number` from the
  workflow event (PR payload on `pull_request`, GH API lookup on `push`).
• Fire-and-forget by default (~15s); heal results arrive via the standard
  Checksum PR-comment pipeline.
• Set `wait: true` to make the action poll status until terminal and exit
  green/red based on the run outcome.

See the README for
the full input/output reference.