Add Compiler Plugin support for the statistics on GroupBy with multiple columns

[zaleslaw]

For next scenarios:

    val res22 = personsDf.groupBy { city }.median("newAge") { age and yearsToRetirement }
    val median221: Int? = res22.newAge[0]
    res22.compareSchemas()

Now it extracts the type from the first column

Partially implemented in #1077

[zaleslaw]

Probably, these functions should be removed in 1.1