FrameColumn.schema can produce types with redundant out variance

Consider FrameColumn where 1 dataframe is empty, and 1 has nullability
Common schema is computed by intersectSchema.

val l = listOf(listOf(3.0))
val col1 = DataColumn.createByType("col", l)
val col2 = col1.take(0)
@Suppress("KotlinConstantConditions")
val col3 = col1.map { it.takeIf { false } }
val df = dataFrameOf(
    "frameCol" to columnOf(
        dataFrameOf(col1), dataFrameOf(col2), dataFrameOf(col3)
    )
)
df.schema().print()

It creates a type with out variance that's redundant here:

frameCol: *
    col: List<out Double>?

df.generateDataClasses().print()

Start with the reproducer in the issue and inspect intersectSchema, which computes the common schema for the empty and nullable DataFrame columns. Run the schema and generateDataClasses examples to confirm the redundant out variance, then trace the schema result to determine the expected type and verify that both outputs no longer contain the unnecessary variance.