Fixed the problem with using the dryGridding() route for CFC computations in serial run. This does involve a re-initialization of the SI tool. If this becomes a problem, check comments in imager_base.py on how to disable and revert back to using a different route for serial run which does not require re-initialization of the SI tool.
What is really required is re-initialization of the CFCache object in
the SI tool, not the whole tool. Ultimately, that's what we should
enable and do.
ATerm, NoOpAterm: Made cacheVBInfo() methods non pure-virtual and moved it's code from EVLAAperture to ATerm.cc. These methods are generic and not telescope specific. This allows psterm=T, aterm=F, wprojplanes>1 to work. imager_base.py: using makeCFCache(), which is driven the same away for parallel and serial runs. Ultimately, C++ side code that runs in serial mode should be deprecated.