Python-graphblas

* performance: task creation for GrB_mxm (saxpy method) didn't
account for any work for A(:,k)*B(k,j) when nnz(A(:,k))==0,
but this takes O(1) work to examine B(k,j). Performance
improvement of up to 10x when nnz(A)<<nnz(B).

* performance: revised swap_rule in GrB_mxm, which decides whether
to compute C=A*B or C=(B'*A')', and variants, resulting in up
to 3x performance gain over v6.1.1 for GrB_mxm (observed;
could be higher in other cases).

* minor revision to AVX2 and AVX512f selection
* cpu_features/Makefile: remove test of list_cpu_features

* added GxB_get options: compiler name and version
* added package: https://github.com/google/cpu_features,
Nov 30, 2021 version
* performance: faster C+=A*B when C is full, A is bitmap/full, and B is
sparse/hyper; added saxpy5 kernel. faster C+=A'*B (dot4 kernel).
* bug fix: deserialization of iso and empty matrices/vectors was broken

bug fix: GrB_Matrix_export; numerical values not properly exported

Versions 6.0.1 and 5.2.1, Nov 27, 2021

* v6.x and v5.x: differ only in GrB_wait, GrB_Info, GrB_SCMP, and GxB_init
* (37) bug fix: C+=A'*B when the accum operator is the same as the monoid
and C is iso-full, and A or B are hypersparse (dot4 method).
* performance: GrB_select with user-defined GrB_IndexUnaryOp ~2x faster.
* performance: faster (MIN,MAX)_(FIRSTJ,SECONDI) semirings