A significant bug using multi-chunked data

Today I have found and fixed a bug that could seriously upset the MAKEMAP "maptol" convergence test when used with data that has been split into more than one chunk. Basically, the map at the end of the first iteration on each new chunk was compared with the final map from the previous chunk, rather than with a blank map full of zeros. The effect of this could be to produce spurious values for the "normalized map change" reported at the end of each iteration, and thus cause the iterative loop to be left prematurely. 

Change to way config files work

With today's cutting-edge starlink, configuration parameters work slightly differently in cases where a value is assigned to a parameter twice. For instance, if you config file holds

450.flt.filt_edge_largescale=600
flt.filt_edge_largescale=500

then previously the value 600 would have been adopted when processing 450 um data. With the new changes, the value 500 will be adopted instead. That is, unqualified values now take priority over qualified values.

This will be most relevant if your config file inherits from another config file (which almost all do), in which parameters are specified with qualifiers. Previously, in order to override such a value in your own config, you needed to include the wavelength. But now, you no longer need to include the wavelength.

These changes have been installed at Hilo. To use them at your locally, you will need to rsync from Hilo.