Drop Old Data

Currently Kapacitor expects data to arrive in order. If data arrives out of order the behavior of the task is undefined. It would be nice to improve Kapacitor to simply drop old data so that writing old data doesn't cause unknown or unwanted behavior.

My initial thoughts are that the `StreamNode` could easily accomplish this.

Start by locating the StreamNode implementation and its existing tests. Trace how out-of-order data currently reaches the task, then define the handling needed for old data to be dropped rather than produce undefined behavior. Done means the behavior is implemented and covered by tests for out-of-order writes.

Written by the indexing model from the issue text.