Might fix the problem Mark noticed where turn-off transients
are erroneously matching the drop that follows startup transients.

If the max scale in the exemplar was a column we weren't using, it
would bail out when looking for that correlation later.  Change things
around so exemplars in RAM only keep around the columns we care about.

This makes it easier to use the filter stuff from other code, but it's
also turning it into more of a spaghetti nightmare.  Might not be
worth continuing down this path.

And adjust the period locator accordingly.
Fitting \sin is the same mathematically, it's just conceptually more
straightforward since we're locating zero crossings anyway.

Just a little less code.  Same results.

This is mostly just intended to make a simple filter example shorter.