GroupBy Jupyter codegen

[Jolanrensen]

Fixes #1221

Uses this PR as its base: #663:

This was a long standing minor inconvenience. Having this variable in a cell1 you couldn't access k in keys in the cell2 cell1:

val groupBy = dataFrameOf("a")("1", "11", "2", "22")
    .groupBy { expr { "a"<String>().length } named "k" }

cell2:

groupBy.keys.k

Other use case is adding a new column to GroupBy

val groupBy = dataFrameOf("a")("1", "11", "2", "22")
    .groupBy { expr { "a"<String>().length } named "k" }
    .add("newCol") { 42 }

groupBy.aggregate { newCol into "newCol" }

But it's now up-to-date and contains some refactoring:

• Refactor of CodeWithConverter -> CodeWithTypeCastGenerator. It was never really a "converter", it can just generate a casting expression for an expression, so...
• I created the TypeCastGenerator interface (the name can still change of course). It has a specific implementation that uses the cast<>() functions from the DataFrame API (instead of having string literals in different places) which also supports multiple type-arguments (for GroupBy).
• I did some additional renaming like markerName -> targetTypeName since it makes it easier to understand without a grasp of the larger context.

[zaleslaw]

I have one question here: do we need the same support for Pivot as for GroupBy? At least create a ticket and make a separate fix? or validate with test

[Jolanrensen]

Actually yes!
We can try to test this by writing val pivot dataFrameOf(...).pivot { ... } and trying to access columns in the cell below.

#1268