Skip using GPG when there's large RA drift. Reset GPG with too many skips.

In looking at GPG guiding, I noticed that when large drifts occur that are not related to periodic error, the GPG remembers them and treat them as periodic anyway. For small drifts this isn't an issue, but for large one-off movements, e.g. due to focusing, this can be a problem. These drifts are remembered, and corrections for them are seen repeated in the future at the multiples of the worm gear's period (falling off after a few periods) causing guiding issues at those times.

To correct this, I simply don't run GPG when the error is very large (> 5 arcseconds), and instead use the standard internal guider's control mechanism. Control returns to GPG once the drift becomes lower than the above threshold. This way the GPG never sees the bad values, and thus the echos don't manifest. If too many samples are skipped, the GPG is reset.

During suspended guiding, if such large drifts are seen, the SNR sent to GPG is artificially set to a very low value so that the GPG reduces its confidence in these values (this didn't seem to work with regular guiding).

The image in this link https://photos.app.goo.gl/PGMJ3PpKbUcSUcMe9 shows some of these echos at the period of my mount (480s).