Add code for a multivariate discrete random variable

[cc7768]

This code is very similar to the code that is in the QE.jl library for DiscreteRV.

We could simply treat multivariate discrete random variables as a univariate discrete random variable (and just make everything long vectors), but that requires working with conversions between linear and cartesian indexing. It also simplifies the interpretation of joint distributions.

This mostly mimics the code from DiscreteRV and adds a call to ind2sub so that it does the conversions for you.

[codecov-io]

Codecov Report

Merging #204 into master will decrease coverage by 0.1%. The diff coverage is 78.57%.

@@            Coverage Diff             @@
##           master     #204      +/-   ##
==========================================
- Coverage   91.12%   91.01%   -0.11%     
==========================================
  Files          24       24              
  Lines        1712     1726      +14     
==========================================
+ Hits         1560     1571      +11     
- Misses        152      155       +3

Impacted Files	Coverage Δ
src/discrete_rv.jl	86.95% <78.57%> (-13.05%)	:arrow_down:

---

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 9334727...d8798d5. Read the comment docs.

[sglyon]

Just throwing this out there -- would it be better to replace the q and Q fields with an instance of DiscreteRV on vec(q)?