scale does not handle all zero arrays properly

[pavlis]

In working on the revision to CRN3CDecon I discovered a bug in our scale function. It did not handle the case where all the sample data were machine zeros. It created a divide by zero = inf problem when applied to that case create arrays of inf values.

There is a fix in the branch I'm working on but it will likely not be merged for a while as I have a lot more testing to do with the revisions to the CNR3CDecon module. My fix was to the C++ code to do nothing if the computed amplitude was a machine 0 and then log a complaint. There is a parallel problem with scale_ensemble I handled almost the same way. That is, if the entire ensemble has nothing but zeros it does nothing. The difference is that at the moment there is a small disconnect in elog with "Core" ensembles. The low-level C++ function template was designed for "Core" ensembles. That problem could be fixed, but it seems unnecessary as something is really wrong with someone's workflow if they try to run with ensembles of all 0s where the member objects are marked live.

I created a set of tests for the revisions and they behave as expected. The only issue is that until I merge this branch this minor bug will be in the distribution.