[SPARK-28858][ML][PYSPARK] add tree-based transformation in the py side

[zhengruifeng]

What changes were proposed in this pull request?

expose the newly added tree-based transformation in the py side

Why are the changes needed?

function parity

Does this PR introduce any user-facing change?

yes, add setLeafCol & getLeafCol in the py side

How was this patch tested?

added tests & local tests

[zhengruifeng]

This PR follows current pyspark's convention, which do support no setter/getter for model.

I am looking into the hierarchy of both scala and py side, and will open a ticket on the sync in next one or two weeks. I think we can sync them in a batch in the future, so in this PR I stick to pyspark's convention.

[srowen]

Looks reasonable to me. I guess unfortunately all of the examples predict a leaf ID of 0?

[zhengruifeng]

@srowen Yes. But I try other inputs and the returned leafId can be 1.

[zhengruifeng]

should I change the input dataset to make some output != 0? @srowen

[srowen]

I wouldn't change the examples unless it's easy. It's not a big deal.

[SparkQA]

Test build #109631 has finished for PR 25566 at commit edc8235.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[BryanCutler]

LGTM

[BryanCutler]

merged to master, thanks @zhengruifeng !

[zhengruifeng]

Thanks for reviewing! @BryanCutler @srowen