Remove objid and thing_id / thingid

[kbarbary]

This removes the SDSS-specific objid and thing_id fields from the core of Celeste (primarily CatalogEntry). Mostly these were not used for anything, so this doesn't have much of an effect.

In the benchmarks, objid was sometimes used as an object index. It was necessary to have an object index when reshaping (melting or stacking) DataFrames. However, it wasn't necessary for the index to be globally unique or meaningful. So here, I've replaced it with a simple running index (1:length(catalog)) within the catalog.

References to objid/thing_id still exist in the SDSSIO module which is totally fine. The code there has the ability to read tables with those fields, they're just not used by any of the rest of the code.

[codecov-io]

Codecov Report

Merging #696 into master will increase coverage by 0.26%. The diff coverage is 76.92%.

@@            Coverage Diff             @@
##           master     #696      +/-   ##
==========================================
+ Coverage   73.89%   74.16%   +0.26%     
==========================================
  Files          36       36              
  Lines        3812     3789      -23     
==========================================
- Hits         2817     2810       -7     
+ Misses        995      979      -16

Impacted Files	Coverage Δ
src/SDSSIO.jl	93.23% <ø> (-0.03%)	:arrow_down:
src/model/light_source_model.jl	53.57% <ø> (ø)	:arrow_up:
src/DeterministicVI.jl	88% <0%> (ø)	:arrow_up:
src/GalsimBenchmark.jl	100% <100%> (ø)	:arrow_up:
src/joint_infer.jl	70.52% <100%> (ø)	:arrow_up:
src/AccuracyBenchmark.jl	43.37% <50%> (+0.08%)	:arrow_up:
src/ParallelRun.jl	40.27% <75%> (+0.27%)	:arrow_up:

---

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 7712bf9...8432045. Read the comment docs.

[jeff-regier]

Great! Bonus: code base is now 56 lines shorter.