Create a separate module for error propagation

[astrochun]

The module would be called error_prop

It will include:

1. Construction of probability distribution functions (PDFs) based on inputs and 1-sigma uncertainties
2. Calculation of 68-percentile uncertainties given a PDF and initial guess.
3. A ratios package that computes different emission-line ratios. This is something that will work with simple emission-line ratios and PDFs of those ratios.
4. Derive properties (T_e, metallicity) from flux ratio PDFs
5. Write npz files, such as that done in Zcalbase_gal
6. Define npz files in column_names module
7. Change metallicity_calculation to return a single dictionary instead of splitting content out
8. Write PDFs for derived properties
9. Write revised derived properties table
10. Set numpy random seeds

[astrochun]

We might want to make construct_pdf() to handle multiple sets. Perhaps use a dictionary approach?

[astrochun]

During group discussion, an idea that I came up with is a function that would generate the flux ratios and derived properties regardless of individual measurements or distribution function.

[astrochun]

Note: composite_indv_detect will need to be updated for metallicity_calculation() return information. This can be done within release/0.6.0. This way, I would not need to merge in composite_indv_detect

[astrochun]

Code works for Caroline's example dataset that was provided. For Reagen, she had to switch to Python 3, but that was due to a dependency fix in column_names module that was not incorporated within feature/error_prop. After the merge of PR #24 and resolving conflicts, we should be good.

[astrochun]

Issue is now resolved.