### What changes were proposed in this pull request?
[SPARK-33546] stated the there are three inconsistency behaviors for CREATE TABLE LIKE.
1. CREATE TABLE LIKE does not validate the user-specified hive serde. e.g., STORED AS PARQUET can't be used with ROW FORMAT SERDE.
2. CREATE TABLE LIKE requires STORED AS and ROW FORMAT SERDE to be specified together, which is not necessary.
3. CREATE TABLE LIKE does not respect the default hive serde.
This PR fix No.1, and after investigate, No.2 and No.3 turn out not to be issue.
Within Hive.
CREATE TABLE abc ... ROW FORMAT SERDE 'xxx.xxx.SerdeClass' (Without Stored as) will have
following result. Using the user specific SerdeClass and fetch default input/output format from default textfile format.
```
SerDe Library: xxx.xxx.SerdeClass
InputFormat: org.apache.hadoop.mapred.TextInputFormat
OutputFormat: org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat
```
But for
CREATE TABLE dst LIKE src ROW FORMAT SERDE 'xxx.xxx.SerdeClass' (Without Stored as) will just ignore user specific SerdeClass and using (input, output, serdeClass) from src table.
It's better to just throw an exception on such ambiguous behavior, so No.2 is not an issue, but in the PR, we add some comments.
For No.3, in fact, CreateTableLikeCommand is using following logical to try to follow src table's storageFormat if current fileFormat.inputFormat is empty
```
val newStorage = if (fileFormat.inputFormat.isDefined) {
fileFormat
} else {
sourceTableDesc.storage.copy(locationUri = fileFormat.locationUri)
}
```
If we try to fill the new target table with HiveSerDe.getDefaultStorage if file format and row format is not explicity spefified, it will break the CREATE TABLE LIKE semantic.
### Why are the changes needed?
Bug Fix.
### Does this PR introduce any user-facing change?
No.
### How was this patch tested?
Added UT and Existing UT.
Closes#30705 from leanken/leanken-SPARK-33546.
Authored-by: xuewei.linxuewei <xuewei.linxuewei@alibaba-inc.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
e7fe92f129