[BUG] test_cast_date_integral fails on databricks 3.4.1

[revans2]

Describe the bug
#9957 added a new test that fails on databricks because not all of the casts are supported on databricks by default.

[2023-12-06T14:15:54.218Z] 
[2023-12-06T14:15:54.218Z] [gw2]�[36m [ 12%] �[0m�[32mPASSED�[0m ../../src/main/python/cast_test.py::test_cast_fallback_not_UTC[SetValues(not_null)(StringType(),['2023-03-20 10:38:50', '2023-03-20 10:39:02'])-TIMESTAMP][DATAGEN_SEED=1701870720, INJECT_OOM, ALLOW_NON_GPU(ProjectExec)] 
[2023-12-06T14:15:54.471Z] ../../src/main/python/cast_test.py::test_cast_date_integral[DATAGEN_SEED=1701870720] 23/12/06 14:15:54 WARN GpuOverrides: 
[2023-12-06T14:15:54.472Z]   ! <RDDScanExec> cannot run on GPU because GPU does not currently support the operator class org.apache.spark.sql.execution.RDDScanExec
[2023-12-06T14:15:54.472Z]     @Expression <AttributeReference> a#25534 could run on GPU
[2023-12-06T14:15:54.472Z]     @Expression <AttributeReference> b#25535 could run on GPU
[2023-12-06T14:15:54.472Z] 
[2023-12-06T14:15:54.725Z] 23/12/06 14:15:54 WARN GpuOverrides: 
[2023-12-06T14:15:54.726Z] !Exec <ProjectExec> cannot run on GPU because not all expressions can be replaced
[2023-12-06T14:15:54.726Z]   @Expression <Alias> cast(a#102050 as tinyint) AS a#102052 could run on GPU
[2023-12-06T14:15:54.726Z]     !Expression <Cast> cast(a#102050 as tinyint) cannot run on GPU because Cast from DateType to ByteType is not supported
[2023-12-06T14:15:54.726Z]       @Expression <AttributeReference> a#102050 could run on GPU
[2023-12-06T14:15:54.726Z]   @Expression <Alias> cast(a#102050 as smallint) AS a#102053 could run on GPU
[2023-12-06T14:15:54.726Z]     !Expression <Cast> cast(a#102050 as smallint) cannot run on GPU because Cast from DateType to ShortType is not supported
[2023-12-06T14:15:54.727Z]       @Expression <AttributeReference> a#102050 could run on GPU
[2023-12-06T14:15:54.727Z]   @Expression <Alias> cast(a#102050 as int) AS a#102054 could run on GPU
[2023-12-06T14:15:54.727Z]     !Expression <Cast> cast(a#102050 as int) cannot run on GPU because Cast from DateType to IntegerType is not supported
[2023-12-06T14:15:54.727Z]       @Expression <AttributeReference> a#102050 could run on GPU
[2023-12-06T14:15:54.727Z]   @Expression <Alias> cast(a#102050 as bigint) AS a#102055L could run on GPU
[2023-12-06T14:15:54.727Z]     !Expression <Cast> cast(a#102050 as bigint) cannot run on GPU because Cast from DateType to LongType is not supported
[2023-12-06T14:15:54.727Z]       @Expression <AttributeReference> a#102050 could run on GPU
[2023-12-06T14:15:54.727Z]   ! <RDDScanExec> cannot run on GPU because GPU does not currently support the operator class org.apache.spark.sql.execution.RDDScanExec
[2023-12-06T14:15:54.727Z]     @Expression <AttributeReference> a#102050 could run on GPU
[2023-12-06T14:15:54.727Z] ```

Not sure if this is something special that was pulled back into databricks to make it more closely match ANSI SQL or what.

[NvTimLiu]

Besides Databricks environment, the same failure also occurred on our nightly spark-3.4.1 & spark-3.5.0 integration tests.

[NVnavkumar]

This failure also occurs in Spark 3.4.0 as well

[res-life]

Fix: #9984