Fix neutron angular distribution for fng_dose

[paulromano]

I realized that the angular distribution of neutrons in the fng_dose model was not quite right. By using discrete values of mu, it was overweighting the poles (close to -1 and 1). This branch changes the way the angle distributions are setup to use uniform distributions over ranges of mu rather than discrete values and then modifies the relative intensities based on the width of the mu intervals.

[eepeterson]

@paulromano were the source distributions specified for MCNP discrete or tabular? Or could you briefly comment on how this source is generated?

[paulromano]

Sorry, let me elaborate a bit. The MCNP model uses a compiled source that actually simulates D-T collisions to produce neutrons, so it is neither discrete nor tabular. Until we are able to have an equivalent custom source in C++ (or a wrapper around the MCNP Fortran source), my approach here has been to use the tabulated data from fngdos-e.htm in the SINBAD distribution, which gives the relative intensity for discrete cos(θ) values, and then for each of those cos(θ) values a tabulated energy distribution. The problem this PR solves is that if I only produce source neutrons at those discrete cos(θ) values, it overweights directions close to the poles. The other thing the PR fixes is the reference direction for neutrons: before it was (0, 0, 1) and this fixes it to (0, 1, 0), i.e., forward in the y-axis. With the fix to the reference direction, the overweighting of cos(θ)=1 was resulting in a dose rate that was too high.

[eepeterson]

OK thanks for the detailed explanation. Digging into the benchmark files and looking through the docs I opened issue #3 to remind me of a potentially useful feature. Otherwise I think this is good to go