Transpose and adjoint solve capabilities for SparseLU

The changes add support for transpose and adjoint solves to SparseLU. Given a sparse matrix A and a right-hand side vector b, additionally to solving A x1 = b, now it is also possible to reuse the factorization of A to also solve the adjoint problem A' x2 = b and the transposed problem A^T x3 = b. An example code is:

VectorXd x1(n), x2(n), x3(n), b(n);

SparseMatrix<double> A;

SparseLU<SparseMatrix<double>, COLAMDOrdering<int> > solver;

// fill A and b;

// Compute the ordering permutation vector from the structural pattern of A

solver.analyzePattern(A);

// Compute the numerical factorization

solver.factorize(A);

//Use the factors to solve the linear system, the adjoint system and the transposed system

x1 = solver.solve(b);

x2 = solver.adjoint().solve(b);

x3 = solver.transpose().solve(b);

The affected files are

Eigen/src/SparseLU/SparseLu.h: providing functionality and documentation for the new member functions adjoint() and transpose(). To this and the two member variables m_transposeView and m_adjointView are added, that increase the size of of a SparseLU instance by a few bytes.

Eigen/src/SparseLU/SparseLU_SupernodelMatrix.h: Add member function solveTransposedInPlace() for adjoint and transpose solve and related code.

test/sparse_solver.h: Tests of the new member functions are added in a form of template function specialization of check_sparse_solving().

I hope that the changes are helpful to some users. I am eager for comments and also willing to improve the code/documentation.