Improve algorithmic complexity of sampling among multiple sources

[paulromano]

Bug Description

If you have multiple Source objects specified for settings.source, OpenMC will sample among them based on the relative strengths of the sources (see the sample_external_source function). Currently, the algorithm for sampling is just a linear table search over the CDF, which means it is O(n) where n is the number of Source objects. We could make this O(1) by using alias sampling, which can be provided by the DiscreteIndex class.

[gonuke]

Is this still available @johvincau ? @paulromano ? This looks like a nice first project for one of my students, if so....

[gonuke]

Steps:

1. add a new "global" variable adjacent to external_sources that is of type DiscreteIndex
2. initialize that discrete index after the external_sources are initialized src/settings.cpp#L560 using a vector of strengths (For s in external_sources s->strength()
3. sample DiscreteIndex in sample_external_sources to get i

[gonuke]

@paulromano - any chance you can reassign this to @magnoxemo, since he's actively working on it?

[paulromano]

I can't assign him since he's not a member on the openmc-dev organization, but for now I left the assignee blank. @gonuke Thanks for letting us know, and thanks @magnoxemo for working on this!