fix: deployment health check, container cleanup, and status reporting

Three fixes for the deployment pipeline:
1. Health check path: /health -> /cameleer/health (matches agent)
2. Container cleanup: stop AND remove old container before starting
   new one, plus orphan cleanup by container name to prevent conflicts
3. Container status: read health.status instead of state.status so
   waitForHealthy correctly detects the "healthy" state

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

src/main/java/net/siegeln/cameleer/saas/deployment/DeploymentService.java

@@ -119,12 +119,23 @@ public class DeploymentService {
                         var oldContainerId = (String) oldMetadata.get("containerId");
                         try {
                             runtimeOrchestrator.stopContainer(oldContainerId);
+                            runtimeOrchestrator.removeContainer(oldContainerId);
                         } catch (Exception e) {
-                            log.warn("Failed to stop old container {}: {}", oldContainerId, e.getMessage());
+                            log.warn("Failed to stop/remove old container {}: {}", oldContainerId, e.getMessage());
                         }
                     }
                 });
             }
+            // Also try removing any container with the same name (handles orphaned containers)
+            try {
+                var existing = runtimeOrchestrator.getContainerStatus(containerName);
+                if (!"not_found".equals(existing.state())) {
+                    runtimeOrchestrator.stopContainer(containerName);
+                    runtimeOrchestrator.removeContainer(containerName);
+                }
+            } catch (Exception e) {
+                // Container doesn't exist — expected for fresh deploys
+            }
 
             // Build Traefik labels for inbound routing
             var labels = new java.util.HashMap<String, String>();

src/main/java/net/siegeln/cameleer/saas/runtime/DockerRuntimeOrchestrator.java

@@ -94,7 +94,7 @@ public class DockerRuntimeOrchestrator implements RuntimeOrchestrator {
                 .withHostConfig(hostConfig)
                 .withHealthcheck(new HealthCheck()
                         .withTest(List.of("CMD-SHELL",
-                                "wget -qO- http://localhost:" + request.healthCheckPort() + "/health || exit 1"))
+                                "wget -qO- http://localhost:" + request.healthCheckPort() + "/cameleer/health || exit 1"))
                         .withInterval(10_000_000_000L)      // 10s
                         .withTimeout(5_000_000_000L)         // 5s
                         .withRetries(3)
@@ -131,8 +131,12 @@ public class DockerRuntimeOrchestrator implements RuntimeOrchestrator {
         try {
             var inspection = dockerClient.inspectContainerCmd(containerId).exec();
             var state = inspection.getState();
+            var health = state.getHealth();
+            var healthStatus = health != null ? health.getStatus() : null;
+            // Use health status if available, otherwise fall back to container state
+            var effectiveState = healthStatus != null ? healthStatus : state.getStatus();
             return new ContainerStatus(
-                    state.getStatus(),
+                    effectiveState,
                     Boolean.TRUE.equals(state.getRunning()),
                     state.getExitCodeLong() != null ? state.getExitCodeLong().intValue() : 0,
                     state.getError());