A ProfileSPDLinDirectThreadSolver object can be constructed to
solve a ProfileSPDLinSOE object. It does this in parallel using
threads by direct means, using the LDL^t variation of the cholesky
factorization. The matrx A is factored one row block at a time using
a left-looking approach. Within a row block the factorization is
performed by NP threads. No BLAS or LAPACK routines are called
for the factorization or subsequent substitution.