Milestones of this project

In fact if we put the parallel stuff aside, the work seems pretty easy to be done. I have roughly read all the docs of the RNGs mentioned in #1 and have some ideas.

TODOs is generally as the followings (in order):
- [x] Complete the design of test suites and main framework of RNGs. #4
- [x] Complete a method of benchmark. It will also used on the current `Base.rand`. #2
- [x] Implement the various RNGs in pure Julia. Make sure that they can pass the tests.
- [x] Dig into the codes for better performance, maybe write some C or llvm codes and even do some changes to Julia itself.
- [ ] Select the RNG of the best performance. Try to merge it to `Base.Random`.
- [ ] Take parallel into account.
- [x] Documentation and result reports etc.

In my opinion it is suitable to work on the first three by the midterm evaluation.

Issues have been created in various aspects, and this is one for general discussion. I'm willing to see any advice.