Implement sampler as independent function

- uniform distribution sampler
- log uniform distribution sampler
  -  The terms in NLP application always meet heavy-tailed distribution. So we need a log uniform distribution sampler. And the  distribution`s PDF  is 1/(x+1)lnR in which R is the range of sampling.

[How to implement log uniform distribution sampler](https://wanghaoshuang.github.io/2017/11/Log-uniform-distribution-sampler/)