All distributions should have RNGs

[agarie]

Some of the currently supported distribution don't have RNGs, e.g. Fisher-Snedecor and exponential.

• Triage the code to see which distributions lack a RNG.
• Prepare a plan to implementing them. Can we copy some implementation or should we roll our own? (this is actually quite easy for the ones we currently support)

[envp]

Currently these are the distribution with an rng method (grep -rl def\ rng ./lib/distribution):

• Chi-squared
• Exponential
• Logistic
• Normal
• Uniform

That leaves the following with no rng method defined:

• Beta
• Binomial
• Bivariatenormal
• F
• Gamma
• Hypergeometric
• Lognormal
• Normalmultivariate
• Poisson
• Student's T
• Weibull

Let me know if I missed something.

[agarie]

@vaibhav-y that's very good. Can you update the table in the README.md file? Just send a PR with that and I'll merge it.

[envp]

PR #34 added for README.

This PR doesn't consider changes uniform distribution since SciRuby/distribution hasn't merged it yet. I will rebase that PR and update the README there itself.

Edit:
Please don't close this PR yet, I intend to use this to track the implementations for the rng methods which seem to be the only outstanding things as of now (apart from gamma distribution's Inverse CDF)

[agarie]

That's fine. I'll merge #34 as soon as it passes Travis.