fix(sentinel): stop hourly check job pile-up and self-heal Sentinel

- Start Sentinel with --restart unless-stopped so Docker restarts it
  after a crash or reboot.
- Make CheckAndStartSentinelJob unique per server so copies cannot
  collect in a slow queue.
- Read X-Sentinel-Version on push and update an outdated Sentinel
  without SSH; skip the hourly SSH version check for servers that
  report their version.

tests/Unit/ServerManagerJobSentinelCheckTest.php

@@ -9,6 +9,7 @@ use App\Models\Server;
 use App\Models\Team;
 use Illuminate\Foundation\Testing\RefreshDatabase;
 use Illuminate\Support\Carbon;
+use Illuminate\Support\Facades\Cache;
 use Illuminate\Support\Facades\Queue;
 use Tests\TestCase;
 
@@ -56,6 +57,15 @@ it('dispatches an hourly Sentinel version check for a healthy Sentinel', functio
     });
 });
 
+it('skips the hourly SSH version check when Sentinel reports its version on push', function () {
+    $server = createSentinelCheckServer($this->team, Carbon::now());
+    Cache::put(Server::sentinelReportedVersionCacheKey($server->id), '1.0.2', now()->addHours(2));
+
+    (new ServerManagerJob)->handle();
+
+    Queue::assertNotPushed(CheckAndStartSentinelJob::class);
+});
+
 it('does not schedule periodic Sentinel restart checks', function () {
     $root = dirname(__DIR__, 2);
     $manager = file_get_contents($root.'/app/Jobs/ServerManagerJob.php');