Cannot express a truncated uniform prior over missing values in BSR

[Víctor de Buen Remiro]

Now missing values accept prior with scalar distribution normal, truncated normal and pure non informative (no prior).
Adding truncated uniform prior we could use it to simulate censored data.
This feature should be used carefully over input missing due it could give a non regular regression matrix if there are simultaneous input missing without normal or truncated normal prior.

[Víctor de Buen Remiro]

(In [1531]) Fixed #758

[Víctor de Buen Remiro]

There are four alternative ways to define scalar priors over missing values in function of fields of structure @Bsr.Missing.Info

1. Truncated Normal:

       Prior.Average is known and finite
   and
       0 < Prior.Sigma < +INF
   and
       Prior.LowerBound > -INF 
     or
       Prior.UpperBound < +INF 
   

2. Normal:

       Prior.Average is known and finite
   and
       0 < Prior.Sigma < +INF
   and
       Prior.LowerBound = -INF or ?
   and
       Prior.UpperBound = +INF or ?
   

3. Truncated uniform:

       Prior.Average is known and finite but is used just as initial value
   and
       Prior.Sigma = +INF or ?
   and
       Prior.LowerBound > -INF 
     or
       Prior.UpperBound < +INF 
   

4. None:

       Prior.Average is known and finite but is used just as initial value
   and
       Prior.Sigma = +INF or ?
   and
       Prior.LowerBound = -INF or ?
   and
       Prior.UpperBound = +INF or ?