feat: Add array_max function support

[erenavsarogullari]

Which issue does this PR close?

Closes #14469.

What changes are included in this PR?

Currently, Spark, Snowflake and Presto support array_max function. This can also be useful for DataFusion.

array_max(make_array(3,1,4,2)) => 4
array_max(make_array('h','e','o','l','l',NULL)) => o
array_max(make_array(NULL,NULL)) => NULL

Spark: https://docs.databricks.com/en/sql/language-manual/functions/array_max.html
Snowflake: https://docs.snowflake.com/en/sql-reference/functions/array_max
Presto: https://prestodb.io/docs/current/functions/array.html#array_max-x-x

All potential use-cases have been covered like different data_types, empty array, NULL etc.

Are these changes tested?

Added new UT cases to verify array_max function in terms of different source arrays.

Are there any user-facing changes?

Yes, new SQL function is supported and documentation has also be updated.

[alamb]

FYI @findepi I think you mentioned this feature recently

[jayzhan211]

I have the same question for array_min, but if this function is highly interested from many people then adding it to datafusion core is not a bad idea.
#14417 (comment)

[erenavsarogullari]

I have the same question for array_min, but if this function is highly interested from many people then adding it to datafusion core is not a bad idea. #14417 (comment)

Thanks @jayzhan211 for the review. I think we have 2 use cases for array_min/max functions:
They need to be supported natively by DataFusion for Comet users as a clear API and abstraction and also, i think they can be useful for DF users (not using Comet) as part of core module. On the other hand, other query engines (Snowflake, Presto, GoogleSQL) also support them natively in addition to Spark. In terms of these, IMHO, they are candidate for the core module. However, it is also hard to know how much demand we will have from the community so both module options are ok for me.

[findepi]

as_list_array works only for List(_)
the LargeList(_) | FixedSizeList(_, _) codepath doesn't work

AFAIR, it's hard to create instance of FixedSizeList in SQL, so it will be difficult to test.
i'd suggest to leave it out from the PR for now.

[findepi]

What is .value(0)??

[findepi]

add a test where query process couple arrays in single array_max call

SELECT input, array_max(input)
FROM (
  SELECT make_array(d-1, d, d-2)
  FROM (VALUES (0), (10), (20), (30), (NULL)) t(d)
)

[jayzhan211]

I have the same question for array_min, but if this function is highly interested from many people then adding it to datafusion core is not a bad idea. #14417 (comment)

Thanks @jayzhan211 for the review. I think we have 2 use cases for array_min/max functions: They need to be supported natively by DataFusion for Comet users as a clear API and abstraction and also, i think they can be useful for DF users (not using Comet) as part of core module. On the other hand, other query engines (Snowflake, Presto, GoogleSQL) also support them natively in addition to Spark. In terms of these, IMHO, they are candidate for the core module. However, it is also hard to know how much demand we will have from the community so both module options are ok for me.

Sounds like Spark crate is a great place for this #5600

[findepi]

Per project guidelines proposal #13706 it feels to me as belong to core.
It would be great to either finalize those guidelines or close the PR and come up with new guidelines. Debating separately for each individual function is counter productive.