Implement "drop column with index" through dropping index before dropping column

## Enhancement

For `alter table xxx drop column xxx` statement, it will drop both the column and its index. It brings some problems (e.g. https://github.com/pingcap/tidb/issues/40192). It's also not online-enough in some cases: if a `NOT NULL UNIQUE` column is in write-only mode, the user cannot actually write any data into the table, because the default value of this column is always inserted and checked as duplicated... (though, sounds not like a big problem).

We could try to implement "drop column with index" in multiple steps: drop the index first and then drop the column. It will fix this problem. It'll also make the routine of dropping column more unified with other implementations.