Grab bag of organization and test cleanups

[kbarbary]

This PR looks big but it makes no algorithmic changes. It is only code organization and test clean ups. (The files with big diffs are either moving code around or reindentation). The most notable feature is that tests now use @testset uniformly and all extraneous print output has been removed or made optional via logging levels. This means that the test output is now actually readable. Example:

$ runtests.jl sdssio sep wcs
Test Summary: | Pass  Total
sdssio        |   10     10
Test Summary: | Pass  Total
sep           |    5      5
Test Summary: | Pass  Total
wcs           |    5      5

Timing info:
                test_sdssio.jl:    3.02 seconds
                   test_sep.jl:    1.09 seconds
                   test_wcs.jl:    8.16 seconds
                    Total time:   12.27 seconds

Being able to read the test output made it possible to notice several deprecation warnings which I've fixed.

Detailed summary of changes:

• Rename SkyPatch to ImagePatch, because it represents a pixel box on a given image, not a region on the sky.
• Code movement: move detect_sources() to top level and own file. Move contents of joint_infer.jl into ParallelRun.jl, except partitioning related code which gets its own file. This organization makes sense for future changes I plan to make to organize the main entry points for joint and single inference: having them in the same file is handy. The partitioning stuff is independent of everything else so it can live by itself.
• Make neighbor_map a dictionary mapping (number in catalog) -> (vector of numbers in catalog). I find this easier to think about and it allows some minor code simplifications.
• Log module: allow multiple logging levels: error, warn, info, debug. Allows one to show detail during debugging or running, but to hide the details when running tests.
• All tests now get their images and catalogs from disk via SampleData.get_sdss_[images,catalog](run, camcol, field) functions. This removes some code duplication and will make future refactoring easier.
• Extraneous printouts in tests and library code removed, moved to scripts or made optional.
• Remove some unused functions from SDSSIO.

[codecov[bot]]

Codecov Report

Merging #725 into master will increase coverage by 0.61%. The diff coverage is 88.49%.

@@            Coverage Diff             @@
##           master     #725      +/-   ##
==========================================
+ Coverage   80.88%   81.49%   +0.61%     
==========================================
  Files          37       38       +1     
  Lines        3965     3934      -31     
==========================================
- Hits         3207     3206       -1     
+ Misses        758      728      -30

Impacted Files	Coverage Δ
src/mcmc/mcmc_misc.jl	33.89% <ø> (ø)	:arrow_up:
src/deterministic_vi/elbo_args.jl	100% <ø> (ø)	:arrow_up:
src/mcmc/mcmc_infer.jl	47.32% <ø> (ø)	:arrow_up:
src/model/fsm_util.jl	100% <ø> (ø)	:arrow_up:
src/model/log_prob.jl	90.57% <ø> (ø)	:arrow_up:
src/DeterministicVI.jl	92.3% <ø> (ø)	:arrow_up:
src/partition.jl	100% <100%> (ø)
src/GalsimBenchmark.jl	100% <100%> (ø)	:arrow_up:
src/mcmc/slicesample.jl	73% <100%> (ø)	:arrow_up:
src/detection.jl	100% <100%> (ø)
... and 11 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 215be92...756f3df. Read the comment docs.

[jeff-regier]

Excellent. These all seem like nice improvements.