[HUDI-4038] Avoid calling getDataSize after every record written (#5497)

- getDataSize has non-trivial overhead in the current ParquetWriter impl, requiring traversal of already composed Column Groups in memory. Instead we can sample these calls to getDataSize to amortize its cost.

Co-authored-by: sivabalan <n.siva.b@gmail.com>

hudi-client/hudi-java-client/src/test/java/org/apache/hudi/table/action/commit/TestJavaCopyOnWriteActionExecutor.java

@@ -379,7 +379,7 @@ public class TestJavaCopyOnWriteActionExecutor extends HoodieJavaClientTestBase
 
     List<HoodieRecord> records = new ArrayList<>();
     // Approx 1150 records are written for block size of 64KB
-    for (int i = 0; i < 2000; i++) {
+    for (int i = 0; i < 2050; i++) {
       String recordStr = "{\"_row_key\":\"" + UUID.randomUUID().toString()
           + "\",\"time\":\"2016-01-31T03:16:41.415Z\",\"number\":" + i + "}";
       RawTripTestPayload rowChange = new RawTripTestPayload(recordStr);
@@ -402,7 +402,8 @@ public class TestJavaCopyOnWriteActionExecutor extends HoodieJavaClientTestBase
         counts++;
       }
     }
-    assertEquals(5, counts, "If the number of records are more than 1150, then there should be a new file");
+    // we check canWrite only once every 1000 records. and so 2 files with 1000 records and 3rd file with 50 records.
+    assertEquals(3, counts, "If the number of records are more than 1150, then there should be a new file");
   }
 
   @Test