Use size_type as return type of SparseMatrix::m().

[bangerth]

This makes these functions consistent with what we have in the Epetra wrappers, and (that was my original motivation) avoids warnings in the block matrix class about comparing integers of different sizes/signedness.

Conceptually, it makes sense to use size_type for these functions. size_type is defined as

include/deal.II/lac/trilinos_tpetra_sparse_matrix.h:      using size_type = dealii::types::global_dof_index;

So it is unsigned, which is exactly what these functions should return.

(I will note as an aside that different parts of the Tpetra wrappers define their own size_type types, and some of these are signed:

include/deal.II/lac/trilinos_tpetra_block_sparse_matrix.h:      using size_type       = typename BaseClass::size_type;
include/deal.II/lac/trilinos_tpetra_block_vector.h:      using size_type       = typename BaseClass::size_type;
include/deal.II/lac/trilinos_tpetra_solver_direct.h:      using size_type = dealii::types::signed_global_dof_index;   ******
include/deal.II/lac/trilinos_tpetra_sparse_matrix.h:      using size_type = dealii::types::global_dof_index;
include/deal.II/lac/trilinos_tpetra_sparse_matrix.templates.h:      using size_type = dealii::types::signed_global_dof_index;  ******
include/deal.II/lac/trilinos_tpetra_sparsity_pattern.h:        using size_type = dealii::types::signed_global_dof_index;  ******
include/deal.II/lac/trilinos_tpetra_sparsity_pattern.h:        using size_type = size_t;  ******
include/deal.II/lac/trilinos_tpetra_sparsity_pattern.h:      using size_type = dealii::types::signed_global_dof_index;  ******
include/deal.II/lac/trilinos_tpetra_vector.h:      using size_type = dealii::types::global_dof_index;
include/deal.II/lac/trilinos_tpetra_vector.h:      using size_type  = types::global_dof_index;

This is perhaps not great design. I'll open an issue about this.)

[bangerth]

@jpthiele @kinnewig FYI