Mask out part of a time series based on another time series

This option is available to be able to compute conditional
relationships. An example would be the ENSO teleconnections to Europe
for years in which the NAO is positive. One first uses this function
to generate a new time series with all winters with negative NAO set
to undefined, so that only the positive ones remain. Next that time
series is correlated to an ENSO index as usual.
Note that conditional correlations depending on the time series itself can be done directly from the correlation form. This option is only needed when the condition involves a third time series.