This option restricts the statistical analysis to a subset of fixels, as defined by a boolean fixel data file.
Calculation of the fixel-fixel connectivity matrix, data smoothing, t-statistic calculation, and stiatistical enhancement, are all only performed on those fixels within the mask.
However, the output fixel directory contains all of the fixels from the input template fixel image, not just those within the mask; this retains fixel-fixel correspondence between the input template image and the output results.
For fixels outside the mask, a value of NaN is written for all fixel data file outputs.

Also includes a bug fix where opting to not smooth the input data would likely have resulted in an error.

On testing prior changes to fixelcfestats to support the -mask option, it was found to run extremely slowly, even when not using the -mask option. The suspected cause was that during the matrix normalisation, each row of the connectivity matrix (a map<uint32_t, float>) was being re-constructed from scratch (due to the possibility of fixel indices changing), and hence what was a balanced binary search tree would become essentially a sorted doubly-linked-list. Now, each row of the connectivity matrix is instead converted to essentially a vector<pair<uint32_t,float>>, since it is no longer necessary to have fast lookup for an arbitrary fixel index. This should reduce memory requirement after matrix construction has completed, and will also hopefully improve cache performance since the connectivity data for each fixel will be serialised.
There are also various little fixes and tweaks to polish off additions for supporting the -mask option.

…option