More precise analysisdata histogram accumulation
Accumulate histograms within a frame as 64-bit integers (if possible) or
as double-precision values to support cases like RDFs where there can be
a very large number of points contributing to the histogram.
Add a supporting class that could be used in other analysis modules as
well to make them capable of parallel processing (at least the
AnalysisDataFrameAverager could be adapted to use this approach).
Also, make the definition of "parallelization factor" the same here
and in datastorage.*, adapting a simpler definition.
Change-Id: I63deed99d3c0d3c168d11ab0d85981f1ed6fee8d