Avoid selecting all guide stars in one quadrant

[taldcroft]

For obsids like 20433 proseco has chosen a guide catalog with all stars in one quadrant. This is acceptable but should be avoided when possible.

[jeanconn]

I'm not as sure about the generalization. For example, I think one star away from the cluster has a potential to increase dr95. If we want to look at putting back the figure-of-merit optimization we can, but I'm was hoping to skip ahead to annie-lite-style optimization.

[jeanconn]

This issue is not (explicitly) fixed by #258 . Should I add another "check" that the stars aren't all in one quadrant as part of the optimization?

[jeanconn]

I'll add this check as a separate PR and then we can decide to merge or not after evaluating impact.

[taldcroft]

Not for 4.4.