Improve documentation of Trilinos LA support for linear operators

[jppelteret]

As recently indicated on the mailing list (re. "Linear operators - TrilinosWrappers"), the documentation on how to use linear operators in conjunction with the Trilinos LA classes is a bit sparse. Perhaps its worth putting in a couple of examples, like what's below, to indicate what's important when setting them up.

From https://github.com/dealii/dealii/blob/master/tests/lac/schur_complement_04.cc

using VectorType      = TrilinosWrappers::MPI::Vector;

// Note that the vector type has to be specified
// Also note that this call to linear_operator is
// actually to TrilinosWrappers::linear_operator
const auto lo_A = linear_operator<VectorType>(A);
const auto lo_B = linear_operator<VectorType>(B);
const auto lo_C = linear_operator<VectorType>(C);
const auto lo_D = linear_operator<VectorType>(D);

SolverControl                        solver_control_A(100, 1.0e-10);
TrilinosWrappers::SolverCG           solver_A(solver_control_A);
TrilinosWrappers::PreconditionJacobi preconditioner_A;
preconditioner_A.initialize(A);
   
const auto lo_A_inv = inverse_operator(lo_A, solver_A, preconditioner_A);
const auto lo_S   = schur_complement(lo_A_inv, lo_B, lo_C, lo_D);

and from https://github.com/dealii/dealii/blob/master/tests/lac/linear_operator_12a.cc

using VectorType = TrilinosWrappers::MPI::Vector;

SolverControl                      solver_control(1000, 1e-12);
SolverCG<VectorType>               solver(solver_control);
TrilinosWrappers::PreconditionSSOR preconditioner;
preconditioner.initialize(system_matrix);

const auto lo_A     = linear_operator<VectorType>(system_matrix);
const auto lo_A_inv = inverse_operator(lo_A, solver, preconditioner);

output = lo_A_inv * system_rhs;
constraints.distribute(output);

Also better document (e.g. in intro to linear operation section) the linear_operator variant that targets preconditioners:

const auto M = ...;
LA::MPI::PreconditionJacobi prec_M;
prec_M.initialize(M);
const auto op_pre = linear_operator<LA::MPI::Vector>(prec_M);

vs

const auto M = ...;
LA::MPI::PreconditionJacobi prec_M;
prec_M.initialize(M);
const auto op_pre = linear_operator<LA::MPI::Vector>(M, prec_M);