AOM_QM: use SIMD for flat matrices and re-enable tests.
When AOM_QM is enabled, by default quantization matrices are flat unless enabled with --enable-qm=1. Re-use existing SIMD functions when a flat matrix is used, so that there is no speed deficit when AOM_QM is enabled. SIMD for the non-flat case is TBC. Change-Id: I1bb8da70d3dd5858dac15099610ddf61662e3d0d