1) Introduction of ncores as extra parameter. 2) Introduction of Calibration as a new function (replacement of .calibration.wrap). 3) removal of var.fc

If there is a unique forecast, extreme limits and tercile limits must be vectors. Now 1-D array is cot accepted. Observations are proven to be a vector or a numeric value.

Corrected bug in labels of tercile/extreme probabilities. Now uses first panel positions for all labels.

Handling of case of insufficient data, following the comment of Deborah Verfaillie. The function now warns (and does no calibration) when 3 or less observation-model pairs are available for training. A new parameter na.fill has been introduced.