Noisy C/N0 slows down IAR

[ljbade]

When a satellite is close to the C/N0 cutoff in the solution filter, it can 'bounce', constantly being added and removed from the solution, which slows down the IAR solver.

Until we manage to fix the noisy C/N0 estimate, we should change the cutoff to have a hysteresis to prevent this bouncing effect.

/cc @fnoble @sashamitelman

[ljbade]

Hmm there appears to already be a sort of hystresis, but both the use and drop values default to 31.0.

This should be improved.

[ljbade]

I notice SNR tends to bounce +/- 10 dBz, so perhaps we should set the drop value to 26? That combined with the count/time logic.

[ljbade]

I also removed the hard threshold check in use_tracking_channel as it was overriding the count logic.

[ljbade]

That didn't seem to improve it much. Perhaps we need to filter the cn0 more