Standard deviation aggregation expression

[kevinzwang]

We'd like to have support for computing the standard dev as an aggregation, something along the lines of

# global
df.agg(col("data").std_dev())

# grouped
df.groupby("name").agg(col("data").std_dev())

There's an algorithm for distributed std dev we can use in order to not need to first compute and broadcast the mean.

[colin-ho]

Closed with #3005