v0.1.3

Monitor resource gains two detection settings:

- failure_threshold — consecutive failures before an incident opens
  (retry-before-alert). Gates the fail cause only; a success resets it.
- max_runtime_s — maximum duration of a single run. Replaces grace_s for the
  overrun deadline only; the silence rule and first-run seed still use grace_s.
  Not supported on http monitors, rejected at plan time.

Routes and alert templates accept two new event types:

- success — fires only on a successful run completion.
- started — fires when a run begins.

Required by the API's Terraform HCL exporter, which already emits
failure_threshold and max_runtime_s. Older providers reject such a file
entirely with "Unsupported argument".