fix(maxun-core): prevent orphaned crawl loops after timeout

Move isAborted flag reset from cleanup() to run() so that in-flight
crawl loops see the abort flag during teardown and exit cleanly,
preventing CPU/RAM leaks from timed-out runs.

Previously, cleanup() reset isAborted = false during teardown, which
accidentally un-aborted active async crawl loops. The loops would then
continue running against a closed Playwright browser context, consuming
CPU/RAM and flooding server logs with errors.

Now, isAborted is only reset at the start of a new run() call, which is
the only time it should be cleared. This covers all three timeout
triggers: task-runner, scheduler, and API record handler.

Refs: #1183

maxun-core/src/interpret.ts

@@ -3223,6 +3223,7 @@ export default class Interpreter extends EventEmitter {
    */
   public async run(page: Page, params?: ParamType): Promise<void> {
     this.log('Starting the workflow.', Level.LOG);
+    this.isAborted = false;
     const context = page.context();
 
     page.setDefaultNavigationTimeout(100000);
@@ -3294,8 +3295,9 @@ export default class Interpreter extends EventEmitter {
       this.namedResults = {};
       this.serializableDataByType = { scrapeList: {}, scrapeSchema: {}, crawl: {}, search: {} };
 
-      // Reset state
-      this.isAborted = false;
+      // Reset state (isAborted is intentionally NOT reset here — it must
+      // remain true so that in-flight crawl loops see the abort flag and exit.
+      // It is reset at the start of run() instead.)
       this.initializedWorkflow = null;
 
       this.log('Interpreter cleanup completed', Level.DEBUG);