Update of GMA code to solve the PPP problem (technically)

The GMA code has been updated to introduce the Chiba-Smith option to address the problem of PPP (Peelle's Pertinent Puzzle/Paradox). To avoid confusion, we will refer to this code, and results obtained using it, as GMAP. This code revision was accomplished with a minimum of intervention to the original version of GMA in order to avoid introducing coding errors. The Chiba-Smith approach was implemented by means of simple renormalization of the experimental absolute errors (square roots of the variances) after reading them in from the input file. We refer to this option as a 'technical' solution to exclude PPP since it is based on the subjective assumption by Chiba and Smith that when experimenters quote absolute total errors these are calculated by multiplying a fractional error (comparable to percent error) by the measured value. Thus, it is supposed that it is the fractional error that actually reflects the accuracy that the experimenter intends to convey to the reader. The PPP problem is a consequence of discrepancies, i.e., the scatter observed for various presumably comparable data obtained by different experimenters that is frequently beyond the quoted errors. We have found that two distinct effects can lead to the presence of PPP in data evaluated by the least-squares method. One effect can be attributed to the different shapes of distinct strongly correlated data sets. We choose to label the PPP effect that results from these strong correlations as maxi-PPP. The second effect arises when there is a spread of data and absolute uncertainties are assigned. Two data points with the same percent uncertainty (same accuracy), but having different values, will then be weighted differently by the least-squares evaluation process. The lowest point will be assigned the heaviest weight since the weighting factor corresponds to the reciprocal square of the absolute error