## What changes were proposed in this pull request?

Continue the work of #12789 to rename ml.asve/ml.load to write.ml/read.ml, which are more consistent with read.df/write.df and other methods in SparkR.

I didn't rename `data` to `df` because we still use `predict` for prediction, which uses `newData` to match the signature in R.

## How was this patch tested?

Existing unit tests.

cc: yanboliang thunterdb

Author: Xiangrui Meng <meng@databricks.com>

Closes #12807 from mengxr/SPARK-14831.