Implemente `StatsBase.wsample` so can sample from action space with certain probabilitly like in a policy

[xiaodaigh]

Currently only rand(action_space) is implemented. wsample would be nice.

[findmyway]

There's a TabularRandomPolicy https://github.com/JuliaReinforcementLearning/ReinforcementLearningAnIntroduction.jl/blob/master/src/extensions/policies/tabular_random_policy.jl . I'm not sure whether it should be put into RLBase or not.

[findmyway]

A more general tabular policy will be added into RLCore soon.

[findmyway]

Added in JuliaReinforcementLearning/ReinforcementLearningBase.jl-Archive#70