Would we ever want to expose subqueries in dataframe in the same way they are done in SQL? For example:

subquery = df2.max("key").where(df1["value"] = df2["value"])
df1.where(df1["key"] > subquery).collect()

would be equivalent to the SQL statement

SELECT * FROM df1 WHERE key > (SELECT max(key) FROM df2 WHERE df1.value = df2.value)