fix: avoid launching extra tasks on ECS Fargate

Currently if we run into a capacity issue on Fargate, we wait and
retry again, but without taking into account that some of the tasks
in a given batch of 10 did actually start. This leads to two issues:

1. More tasks than the value of `--count` end up getting launched
2. The check for whether all workers started up fails because the
   number of "ready" files does not match the count, and the test
   run eventually fails with a "sync timeout" error

packages/artillery/lib/platform/aws-ecs/legacy/run-cluster.js

@@ -1713,6 +1713,19 @@ async function ecsRunTask(context) {
 
     try {
       const runData = await ecs.runTask(params).promise();
+
+      const launchedTasksCount = runData.tasks?.length || 0;
+      tasksRemaining -= launchedTasksCount;
+
+      if (launchedTasksCount > 0) {
+        const newTaskArns = runData.tasks.map((task) => task.taskArn);
+        context.taskArns = context.taskArns.concat(newTaskArns);
+        artillery.globalEvents.emit('metadata', {
+          platformMetadata: { taskArns: newTaskArns }
+        });
+        debug(`Launched ${launchedTasksCount} tasks`);
+      }
+
       if (runData.failures.length > 0) {
         artillery.log('Some workers failed to start');
         const uniqueReasons = [
@@ -1723,19 +1736,6 @@ async function ecsRunTask(context) {
         await sleep(10 * 1000);
         throw new Error('Not enough ECS capacity');
       }
-
-      if (runData.tasks?.length > 0) {
-        const newTaskArns = runData.tasks.map((task) => task.taskArn);
-        context.taskArns = context.taskArns.concat(newTaskArns);
-        artillery.globalEvents.emit('metadata', {
-          platformMetadata: { taskArns: newTaskArns }
-        });
-        debug(`Launched ${launchCount} tasks`);
-        tasksRemaining -= launchCount;
-        await sleep(250);
-      } else {
-        retries++;
-      }
     } catch (runErr) {
       if (runErr.code === 'ThrottlingException') {
         artillery.log('ThrottlingException returned from ECS, retrying');