Implement profiling for likelihood scans

[ClarkMcGrew]

Possible a simple addition. The minimizer would need to be called with a fixed parameter for every step of the profile scan. Since that could cause the number of likelihood calculations to explode, it might be necessary to add a better way to specify which parameters to scan (not scan all of them).

[nadrino]

What do you mean exactly? For each point of a given parameter scan, we would throw all the other parameter? I'm not sure I understand this

[ClarkMcGrew]

I mean the usual "frequentist" fix a parameter, minimize the likelihood against the other parameters. The other parameters aren't thrown, they are optimized.

[ClarkMcGrew]

BTW: I haven't started working on this because I wanted to leave it until after validation is finished, and I think it's better tackled with a more generalized "looping" method (e.g. #602)