psastro return status must be accurate

[Paul Price]

The return status from psastro must reflect the result of the astrometry.

When to return a failed exit status can be a complicated question: we don't want to die just because a single chip failed because we can use a model to tie down a few bad chips using the good astrometry from the rest. However, I think we must insist that psastro fails when the final astrometry is bad. To do otherwise is to deceive the pipeline, and to push problems downstream that properly belong in psastro.

psastro should fail under the following circumstances:

• The camera consists of a single chip, which fails to find a solution (I think I've already implemented this case).
• The camera consists of multiple chips which all fail to find a solution.
• The camera consists of multiple chips, a small subset fail to find a solution but there's no model available to apply.
• The camera consists of multiple chips, a large subset fail to find a solution.
• The camera consists of multiple chips, some subset of chips have a solution, but the solutions are discordant.

There should be a note made in the database if the astrometric solution for an exposure required the use of a model to get a solution for some chips.