Validate_batch function for debugging batch calls

[dougbrn]

Currently, batch fails if any lightcurve passed to the analysis function fails. This all or nothing behavior creates some extreme user-unfriendliness at scale, where a single nan can crash a full computation at the very end. Dask error logging is very unsupportive for this as well, as it really doesn't give enough to go on to debug properly.

This motivates a potential debug function, that does batch but guarantees that the run will execute, but the result is all object and/or source rows that failed the computation. This would allow for users to easily key into the cause of the issue between their batch function and the lightcurves and filter/adjust as needed.

[dougbrn]

This is related to #354, and the problem generating this may be solved by the result of it