Failed Cadence submission should have a path

[jchate6]

When a cadenced observation is scheduled for the future, but the target is not visible or submission fails for any reason, we should have a default and user defined path for behavior.

Proposed Default paths:

1. Immediately kill cadence, send message to user
2. Treat failed submission as failed observation, attempt to schedule N more out. If still failed, kill cadence, send message to user.

If there is a user supplied method, run that instead.

[jchate6]

@moira-andrews @cmccully please supply additional information, concerns, details, and nuance.

[moira-andrews]

I think that having a method where the cadence may attempt to resubmit with more flexible observational parameters would be useful over immediately killing the sequence (increasing airmass, lower lunar distance), which could be user defined. This may require more careful thought on how cadence strategies interact with ObservationGroup and ObservationRecord, since our method of modifying sequences involves canceling them and creating a new group, record, and dynamic cadence with the updated parameters.