filter unused posteriors and globals when noob = 0 in oobstats output

[fradav]

When specifying noob 0 we have nonsensical outputs like :

Global (prior) errors
Computed from the mean taken as point estimate
                     NMAE : 0.8594806933131371
                      MSE : 13416.575836132817
                     NMSE : 105.81804491460004
Computed from the median taken as point estimate
                     NMAE : -nan         
                      MSE : -nan         
                     NMSE : -nan         
Confidence interval measures
             90% coverage : -nan         
              Mean 90% CI : -nan         
     Mean relative 90% CI : -nan         
            Median 90% CI : 0            
   Median relative 90% CI : 0            

Local (posterior) errors
Computed from the mean taken as point estimate
                     NMAE : 0.9325835661458552
                      MSE : 16374.299814890093
                     NMSE : 142.17663666951591
Computed from the median taken as point estimate
                     NMAE : -nan         
                      MSE : -nan         
                     NMSE : -nan         

[fradav]

fixed in a80949e