Reduced order methods

[donovan97]

Pull request to add reduced-order methods POD-Galerkin and POD-Petrov Galerkin, with test on Burger's Rewienski.

The changes are the following:

• Addition of PODGalerkinODESolver and PODPetrovGalerkinODESolver classes, derived from ODESolverBase.
• Addition of 3 new factory methods in ODESolverFactory. One factory takes only a DGBase and can create an ImplicitODESolver or an ExplicitODESolver. One factory takes a DGBase and POD and can create a PODGalerkinODESolver and PODPetrovGalerkinODESolver. The other two factories are identical except they also take a ODESolverEnum allowing the user to specify exactly which solver is requested (instead of choosing value specified in the parameter).
• pod_basis with POD is changed to a TrilinosWrapper::SparseMatrix to interface with the ODE solver classes.
• Creation of ReducedOrder test which first runs 3 snapshots. This is done within the Cmake file, by calling the BurgersRewienskiSnapshot test. The ReducedOrder test uses the snapshots to generate a POD basis. All basis functions are used to ensure that the reduced order solution matches the full order solution exactly. If solutions do not match, test fails.

[donovan97]

Tests have been run, all required tests pass.

[donovan97]

Looking good. Would it be possible to provide references to some of the papers you might have used for the POD basis? For example, whenever some code closely related to a formula of a paper, mention paper and formula number.

Done in the most recent commit.