Check solver residuals for station-fetch runs with many stations

[nwagenbrenner]

Occasionally the solver diverges when many stations are used. We should watch the residuals more carefully and alert the user if this happens.

[nwagenbrenner]

This is currently a problem for Wesley. He's doing thousands of simulations and occasionally a run diverges (usually when there are > 3 stations). We should try to determine when the solver is not going to converge and skip to the next run. At the end of the simulation we can report which runs didn't converge. Maybe for a quick fix for Wesley we can just set the number of solver tries to 10 (it's currently at 150)?

[nwagenbrenner]

I see now that NINJA_POINT_MAX_MATCH_ITERS can be set at runtime. Wesley will use that for his work. If we can't match stations in a given run, we just continue to the next one, so that's good. We should probably report which time steps didn't converge at the end of the simulation.

[nwagenbrenner]

related to db2cada