Fix #640: Jupyter integration conflicts with variable type converters from other integrations

[ark-1]

Fixes #640

[koperagen]

It was done on purpose, see the test. I think this should be fixed on kernel side, if possible. It's not obvious why when our function returns null no other converters are called

dataframe/core/src/test/kotlin/org/jetbrains/kotlinx/dataframe/jupyter/CodeGenerationTests.kt

Line 16 in 0f49f72

fun `Type erased dataframe`() {

[ark-1]

I will discuss changing this behavior on the kernel side. Still, your case is already supported by a different API, so I suggest changing it nevertheless.

[koperagen]

Cool :) Then, run assemble and add generated sources to the commit please. No other questions from me

[Jolanrensen]

Yes, we need to make sure this #401 still works

[Jolanrensen]

All tests pass, also the ones introduced by #401, so should be good to go :)

[ark-1]

The problem with the proposed code that it didn't guarantee the ordering of checks and conversions.
I have rewritten the code to use full FieldHandler API, that has a separate accepts method.