implement gemm for the tridiagonal solver

[rasolca]

Depends on part of #868.

Implement gemm such that it can handle the following use case:

• tridiagonal eigensolver(matrix)
  • ...
  • merge(submatrix_reference(matrix))
    • ...
    • task _to_schedule_gemms(subpipeline_matrix(submatrix_reference), Sender_of_gemm_sizes)
      • create submatrix references
      • gemm(set_submatrix_reference_1)
      • gemm(set_submatrix_reference_2)

Work completed for:

• local implementation
  • Enable Gemm (local) to be used with MatrixRef #969
• distributed implementation
  • Enable Gemm (distributed) to be used with MatrixRef #1022