[SPARK-12552][FOLLOWUP] Fix flaky test for "o.a.s.deploy.master.MasterSuite.master correctly recover the application"

## What changes were proposed in this pull request?

Due to the RPC asynchronous event processing, The test "correctly recover the application" could potentially be failed. The issue could be found in here: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/78126/testReport/org.apache.spark.deploy.master/MasterSuite/master_correctly_recover_the_application/.

So here fixing this flaky test.

## How was this patch tested?

Existing UT.

CC cloud-fan jiangxb1987 , please help to review, thanks!

Author: jerryshao <[email protected]>

Closes apache#18321 from jerryshao/SPARK-12552-followup.

(cherry picked from commit 2837b14)
Signed-off-by: Wenchen Fan <[email protected]>

Author: jerryshao

core/src/test/scala/org/apache/spark/deploy/master/MasterSuite.scala

@@ -169,7 +169,7 @@ class MasterSuite extends SparkFunSuite
       master.rpcEnv.setupEndpoint(Master.ENDPOINT_NAME, master)
       // Wait until Master recover from checkpoint data.
       eventually(timeout(5 seconds), interval(100 milliseconds)) {
-        master.idToApp.size should be(1)
+        master.workers.size should be(1)
       }
 
       master.idToApp.keySet should be(Set(fakeAppInfo.id))