[SPARK-45588][SPARK-45640][SQL][TESTS][3.4] Fix flaky ProtobufCatalystDataConversionSuite

### What changes were proposed in this pull request?
The pr aims to fix flaky ProtobufCatalystDataConversionSuite, include:
- Fix the type check (when the random value was empty array, we didn't skip it. Original intention is to skip default values for types.) [SPARK-45588]
- When data.get(0) is null, data.get(0).asInstanceOf[Array[Byte]].isEmpty will be thrown java.lang.NullPointerException. [SPARK-45640]

Backport above to branch 3.4.
Master branch pr: apache#43424 & apache#43493

### Why are the changes needed?
Fix flaky ProtobufCatalystDataConversionSuite.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
- Pass GA
- Manually test

### Was this patch authored or co-authored using generative AI tooling?
No.

Closes apache#43520 from panbingkun/branch-3.4_SPARK-45640.

Authored-by: panbingkun <[email protected]>
Signed-off-by: Jungtaek Lim <[email protected]>

Author: panbingkun

connector/protobuf/src/test/scala/org/apache/spark/sql/protobuf/ProtobufCatalystDataConversionSuite.scala

@@ -136,7 +136,8 @@ class ProtobufCatalystDataConversionSuite
       while (
         data != null &&
         (data.get(0) == defaultValue ||
-          (dt == BinaryType &&
+          (dt.fields(0).dataType == BinaryType &&
+            data.get(0) != null &&
             data.get(0).asInstanceOf[Array[Byte]].isEmpty)))
         data = generator().asInstanceOf[Row]