Use R RNG (random number generator) instead of GLS

[mhunter1]

This says not to use GSL's or non-R RNGs: https://cran.r-project.org/doc/manuals/r-release/R-exts.html

The source version of dynr uses an RNG for ensemble forecasting. Update this to use R instead.

Here's info on the interface: https://cran.r-project.org/doc/manuals/r-release/R-exts.html#Random-numbers