Generalize evals infra to support more types of evals, organization and queuing of named suites (#24941)

evals/interactive-hang.eval.ts

@@ -8,6 +8,8 @@ describe('interactive_commands', () => {
    * intervention.
    */
   evalTest('USUALLY_PASSES', {
+    suiteName: 'default',
+    suiteType: 'behavioral',
     name: 'should not use interactive commands',
     prompt: 'Execute tests.',
     files: {
@@ -49,6 +51,8 @@ describe('interactive_commands', () => {
    * Validates that the agent uses non-interactive flags when scaffolding a new project.
    */
   evalTest('ALWAYS_PASSES', {
+    suiteName: 'default',
+    suiteType: 'behavioral',
     name: 'should use non-interactive flags when scaffolding a new app',
     prompt: 'Create a new react application named my-app using vite.',
     assert: async (rig, result) => {