Follow up because Cmake did not bother to rebuilt a C++ executable that depends
on the header file where I replaced the macros.

This should avoid the awkward handling.  The map makes sure that only the
central frequencies are passed around which have been used in the observation.

Also be nice an handle exceptions thrown by DAL.

I am afraid that the 0Hz central frequency can be attributed to an integer
divided by a double.  This should at least remove the uncertainty.