Null treatment of `sort` does not follow DataFusion's default

[nuno-faria]

Describe the bug

Using df.sort(col) orders nulls first, while DataFusion's default is to order nulls last (see datafusion.sql_parser.default_null_ordering). Ordering nulls last is also the behavior of sort_by and ctx.sql("... order by ...").

To Reproduce

import datafusion

ctx = datafusion.SessionContext()

ctx.sql("create table t (a int)").collect()

ctx.sql("insert into t values (1), (null), (3), (null), (2)").collect()

ctx.table("t").sort(datafusion.Expr.column("a")).show()
ctx.table("t").sort_by(datafusion.Expr.column("a")).show()
ctx.sql("select * from t order by a").show()

DataFrame()
+---+
| a |
+---+
|   |
|   |
| 1 |
| 2 |
| 3 |
+---+
DataFrame()
+---+
| a |
+---+
| 1 |
| 2 |
| 3 |
|   |
|   |
+---+
DataFrame()
+---+
| a |
+---+
| 1 |
| 2 |
| 3 |
|   |
|   |
+---+

Expected behavior
Follow the same behavior as the default.