1. The exponential power distribution method could be speeded up by
using a rational function approximation for the rejection scaling
parameter.

x2. Check for off by one error in the shuffle algorithm. DONE. checked
against Knuth.

x3. If the user runs gsl-dist for two different distributions then the
variates will be correlated because they come from the same stream of
random numbers. DONE, the user now has to supply a seed

4. Do something about the possibility of the user providing invalid
parameters (e.g. negative variance etc)

5. Add the triangular distribution.

x6. Add the Rayleigh distribution. DONE

7. fix error handling in discrete.c to use GSL_ERROR(), do the same
for memory management

8. Look at Marsaglia & Tsang, "The Monte Python Method for generating
random variables", ACM TOMS Vol 24, No 3, p341