[Kst] NaNs by default in vectors ?

[Brisset, Nicolas]

OK, forget it. It was my mistake: 1.0/0.0 does not yield NaN but Inf... 
Kst does the right thing here...
> Could someone please confirm that vectors are allocated and 
> filled with "NaN" values before being passed to the 
> datasource ? I thought this was the behavior but I see a 
> weird case where values not provided by the datasource are 
> set to "inf", which screws up scaling a bit...