Int32 or Int64 for discrete?

[lucianopaz]

Through a recent discourse thread I found that Discrete allows int16 or int64 dtypes. Given what we had done for parameters regarding #2366 (we set int16 if floatX is float32 and int32 if floatX is float64), should we change Discrete to work with int16 and int32? What do we do with int64's?

[ricardoV94]

Related to #4553

[ferrine]

Every dtype makes sense, can we specify it with dtype arg? I remember struggling with that and fixed in v3, but not v4

[ricardoV94]

Let's not coerce dtype of integer parameters then

[ricardoV94]

Should be closed by #7114