Bugfix for NPE with fillNulls

[Jolanrensen]

fix for #136

The problem with the example was the following:

@DataSchema
interface SchemaA { val i: Int? }

@DataSchema
interface SchemaB { val i: Int }

val df = dataFrameOf("i")(1, null)

df.fillNulls(SchemaA::i).with { it: Int? -> 42 }

// breaks because `null != Int`
df.fillNulls(SchemaB::i).with { it: Int -> 42 } 

This makes sense from a programming standpoint: Essentially a column with null values is matched to a schema with a non-null column. However, since the intention is to fill in those nulls, it might be okay for us to assume that this is fine. I think it's reasonable to say that for any fillNulls- and fillNA call, the values should be considered nullable in the with statement. This PR fixes that.

So now:

df.fillNulls(SchemaA::i).with { it: Int? -> 42 }

df.fillNulls(SchemaB::i).with { it: Int? -> 42 }