fix: scale replicas in response, K8s metrics client, quota precheck, auth tests

- Add GetMetrics method to MetricsClient interface and implement cluster metrics API
- Add QuotaPrecheck service for validating resource quotas before deployment
- Add auth DTO with role/permission models and auth handler tests
- Add instance diagnostics: mounted NFS volumes, labels, annotations in pod diagnostics
- Update workspace handler with GetWorkspace endpoint and shared-user list
- Fix monitoring handler to use correct service method name
- Add tail_lines fallback in instance handler for snake_case query params
- Update nginx config for SSE log streaming support (no buffering)
- Add comprehensive test coverage: auth_service_test, auth_handler_test,
  auth_dto_test, metrics_client_test, quota_precheck_test
- Update error messages for quota validation and instance operations
- ModifyModal: fix YAML lineWidth:0, modified keys summary, delta-only submit
- InstanceCard: correctly disable scale-minus when replicas <= 0
- SidebarLayout: add hover transition for sidebar items
- Update todo.md and lessons.md with latest fixes

backend/internal/adapter/output/k8s/metrics_client_test.go

@@ -0,0 +1,29 @@
+package k8s
+
+import (
+	"testing"
+
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/api/resource"
+)
+
+func TestContainerAllocationCountsVendorGPUMemoryKey(t *testing.T) {
+	container := corev1.Container{
+		Resources: corev1.ResourceRequirements{
+			Requests: corev1.ResourceList{
+				corev1.ResourceName("nvidia.com/gpumem"): resource.MustParse("10000"),
+			},
+			Limits: corev1.ResourceList{
+				corev1.ResourceName("nvidia.com/gpumem"): resource.MustParse("12000"),
+			},
+		},
+	}
+
+	allocation := containerAllocation(container)
+	if allocation.GPUMemoryRequestsMB != 10000 {
+		t.Fatalf("expected GPU memory requests 10000 MB, got %d", allocation.GPUMemoryRequestsMB)
+	}
+	if allocation.GPUMemoryLimitsMB != 12000 {
+		t.Fatalf("expected GPU memory limits 12000 MB, got %d", allocation.GPUMemoryLimitsMB)
+	}
+}