[SPARK-31025][SQL] Support foldable CSV strings by schema_of_csv

[MaxGekk]

What changes were proposed in this pull request?

In the PR, I propose to change checking of the input parameter in the SchemaOfCsv expression, and allow foldable child expression.

Why are the changes needed?

To improve user experience with Spark SQL. Currently, only string literals are acceptable as CSV examples by schema_of_csv:

spark-sql> select schema_of_csv(concat_ws(',', 0.1, 1));
Error in query: cannot resolve 'schema_of_csv(concat_ws(',', CAST(0.1BD AS STRING), CAST(1 AS STRING)))' due to data type mismatch: The input csv should be a string literal and not null; however, got concat_ws(',', CAST(0.1BD AS STRING), CAST(1 AS STRING)).; line 1 pos 7;
'Project [unresolvedalias(schema_of_csv(concat_ws(,, cast(0.1 as string), cast(1 as string))), None)]
+- OneRowRelation

Does this PR introduce any user-facing change?

Yes, after change the schema_of_csv accepts foldable expressions, for example:

spark-sql> select schema_of_csv(concat_ws(',', 0.1, 1));
struct<_c0:double,_c1:int>

How was this patch tested?

• By existing test suites CsvFunctionsSuite and CsvExpressionsSuite.
• Added new test to CsvFunctionsSuite to check foldable input.

[HyukjinKwon]

Max, let's fold from_csv, from_json, and this PRs in single PR. I think it's easier to discuss, review and merge.

[SparkQA]

Test build #119250 has finished for PR 27777 at commit d4da235.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #119288 has finished for PR 27777 at commit 222a31c.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.