v0.29.0

Added

• POST /wp-json/deckwp/v1/site-health — new REST route that
  runs every registered WP_Site_Health test (core + plugin-
  registered via the site_status_tests filter) and returns a
  flat envelope the dashboard's HealthRun model can store.

  Body: empty {}.

  Response 200 (truncated):

  {
    "sent_at": 1717684800,
    "wp_version": "6.9.4",
    "php_version": "8.3.10",
    "summary": { "good": 14, "recommended": 3, "critical": 1, "error": 0 },
    "checks": [
      {
        "test": "wordpress_version",
        "category": "direct",
        "label": "Your version of WordPress is up to date",
        "status": "good",
        "badge_label": "Performance",
        "badge_color": "blue",
        "description": "You are currently running...",
        "actions": ""
      },
      ...
    ]
  }

  Implementation lives in SiteHealth\Runner — instantiates the
  WP_Site_Health singleton, walks both the direct and async
  test buckets returned by get_tests(), and invokes each via
  the get_test_<key> convention on the singleton. Async tests
  whose backing method isn't on the singleton (plugin-registered
  asyncs with custom REST callbacks) are silently skipped — we
  can't safely replay arbitrary plugin REST routes from here.

  Per-test try/catch isolates failures; a broken test surfaces as
  status: 'error' rather than aborting the snapshot. HTML in the
  WP_Site_Health description + actions fields is stripped via
  wp_strip_all_tags — the dashboard renders plain text, and
  shipping arbitrary HTML over the wire only to escape it
  client-side is wasted bandwidth + an XSS landmine.

  HMAC-verified by HmacVerifier. Sets set_time_limit(60) at
  request entry — Site Health includes network-dependent tests
  (loopback, dotorg communication, REST API availability) that
  can each take 5-10s; worst-case sequential cost is ~30-45s on
  a healthy install. The 60s ceiling matches the dashboard's
  outbound HTTP timeout for this trigger.

Wire contract change

Purely additive. Pre-v0.29.0 connectors 404 the new endpoint; the
dashboard's RemoteSiteHealthTrigger (companion commit on the
dashboard repo) surfaces that as an operator-facing
"connector too old — update to v0.29.0+" message.