Fix references to state space of FSP solutions

[dmitrisvetlov]

Within SSIT.m, there are 15 references to fspSoln.stateSpace (among 76 total references to stateSpace).

In at least one place (line 3676, within likelihoodSweep), this causes an error because of the recent refactoring of solution storage.

In other places (namely, in the code related to model reduction), an assignment is made to a local fspSoln.stateSpace, which seems fine a priori, but I have not tested those methods.

[dmitrisvetlov]

@Munsky, on the branch for this issue, I have fixed likelihoodSweep and verified that the fix works in some situations, although it can only be guaranteed to work as written if calling obj.solve() will necessarily create an output model whose Solutions is a 1x1 struct with a stateSpace field.

I am not sure what will happen if the solution method has been set to something other than FSP. Architecturally, a cleaner long-term solution (for this and other problems) is to have the solve method return a solution object that is distinct from the model, and different solution classes will have the appropriate variables.

Note that this also means that likelihoodSweep is not being tested. This can probably be added to some existing test case that validates creation of various visualizations.