fix(deploy): resolve online-eval step on re-deploys with no new configs

The TUI deploy screen renders an "Enable online evaluation" step whenever
projectSpec.onlineEvalConfigs is non-empty, but the step was only
transitioned to a terminal status inside the `if (newOnlineEvalFullSpecs
> 0 ...)` branch. On a re-deploy where every configured online eval was
already deployed in a prior run, that branch is skipped and the step
stays in 'pending', so areStepsComplete() never returns true and the
deploy phase never reaches 'complete' — the TUI hangs after the AWS
deploy itself succeeds.

Mirror the dataset-sync block above by marking the step success when
there are configs but nothing new to enable.

Author: notgitika

src/cli/tui/screens/deploy/useDeployFlow.ts

@@ -656,6 +656,10 @@ export function useDeployFlow(options: DeployFlowOptions = {}): DeployFlowState
         logger.endStep('error', message);
         setOnlineEvalStep(prev => ({ ...prev, status: 'error', error: message }));
       }
+    } else if (onlineEvalFullSpecs.length > 0) {
+      // Step is rendered whenever onlineEvalConfigs is non-empty, but only runs for newly
+      // deployed configs. With nothing new to enable, mark it terminal so the deploy completes.
+      setOnlineEvalStep(prev => ({ ...prev, status: 'success' }));
     }
 
     // Config bundles are now managed via CloudFormation; their state is parsed