Implement accurate and flexible fiberloss calculations

[dkirkby]

This is not ready to merge yet but is getting close. Here is the final checklist:

• Add support for point-like sources.
• Anti-alias fiber aperture.
• Read and use file of optional achromatic centroid offsets.
• Add option to save pre-tabulated fiberloss for later use w/o galsim.
• Review documentation of new features.
• Compare with fiberloss files in desimodel.
• Optimize numerical parameters num_wlen, num_pixels.
• Add some unit tests of new features.

The goal of this PR is to obtain correct results at the required level of realism. Optimizations for efficient calculation of a large number (~5K) of sources distributed over the focal plane in a single exposure are not considered here.

[dkirkby]

The new fiberloss module can now save the results of an on-the-fly galsim calculation to a file with the same format as used in desimodel, so we have a mechanism to update the files desimodel/data/throughput/fiberloss-xxx.dat to be more realistic, if we want to.

[dkirkby]

I am going to close this now (and defer the last 3 items in the checklist) in order to have a stable base for working on #6.