MatrixCSR operations: matvec, transpose, matmul

[chrisrichardson]

Describe new/missing feature

Add new parallel functionality to MatrixCSR for:

• matrix transpose
• matrix-vector product y = Ax, and y = A^T x
• matrix-matrix product C = AB
• matrix triple product P^T A P
• support for matrix transpose, matvec with fixed block size.
• support matrix block size in matmul