Add logcdf to distributions

[mborland]

From #303 and complements #762.

[jzmaddock]

Just checking through to see which others could be usefully specialized,

exponential distribution would be log1p(- exp(-x * λ) )
Geometric: log1p(-pow(1-p, floor(x)))
Logistic: -log1p(exp(-(x-u)/s))
Pareto: log1p(-pow(beta/x, alpha))
Rayleigh: log1p(-exp(-x^2/2)*sigma^2)
Weibull: log1p(-exp(pow(-x/beta, alpha)))

I've just typed those straight in, they need careful checking! :)

[mborland]

@jzmaddock Now that this is green could you take another look through?

[jzmaddock]

Many thanks Matt - this all looks good to go to me - feel free to merge away.

[mborland]

@jzmaddock I added the complement logcdf to the extreme value distribution and the laplace distribution since I missed the latter as well. If you'd like to review those two I can merge once this finishes.