Incorrect star expansion logic MERGE INSERT * / UPDATE *

The natural expectation of INSERT * or UPDATE * in MERGE is to assign source columns to target columns of the same name. However, the current logic here generates the assignment by position. https://github.com/apache/spark/commit/7cfd589868b8430bc79e28e4d547008b222781a5#diff-ed19f376a63eba52eea59ca71f3355d4495fad4fad4db9a3324aade0d4986a47R1214

This can very easily lead to incorrect results without the user realizing this odd behavior.