`aggregateFor` does not use `default` for statistics

[Jolanrensen]

Discovered in #2102 (comment)

val df = dataFrameOf(
    columnOf("Alice", "Bob", "Charlie", "Charlie") named "name",
    columnOf(null, null, null, null) named "group",
)

df.meanFor { group default 0 }.toDataFrame()

The result is NaN even though a default is set.
It turns out that this default is never used for DataFrame.xFor {} calls, as a statistic function can never return null, which is where the default would be used. Probably this default is only used for Pivot(GroupBy) calls.

We should either:

• remove this default call
• Make it work on NaN too (might require some changes in aggregators)
• Make it apply the default before running the statistic

The KDocs will also need to be updated accordingly