r.univar: parallelization for extended statistics

[aaronsms]

This PR addressed #2564, completing parallelization for extended statistics.

The key idea is to maintain "local buckets" for each thread to sort their cells according to the different zones before transfering to the global buffer for computing the order statistics.

In the case of nprocs=1, care is taken to ensure no major overhead, as the first transfer would just be a redirection to the local bucket.

Performance benchmark still needs to be done.

[petrasovaa]

I didn't test yet, but you need to remove the restriction that prevented r.univar -e to run in parallel:
https://github.com/OSGeo/grass/blob/main/raster/r.univar/r.univar_main.c#L155

[petrasovaa]

Weird, the tests in CI work, I am getting locally a lot of errors, e.g.:

GRASS nc_spm_08_grass7/user1:testsuite > r.univar elevation zones=basin_50K -e nprocs=3
WARNING: ZLIB decompression error -3: data error
WARNING: ZLIB decompression error -3: data error
ERROR: Error uncompressing fp raster data for row 0 of <elevation>: error
       code -1
Segmentation fault (core dumped)

[aaronsms]

It did not

Weird, the tests in CI work, I am getting locally a lot of errors, e.g.:

GRASS nc_spm_08_grass7/user1:testsuite > r.univar elevation zones=basin_50K -e nprocs=3
WARNING: ZLIB decompression error -3: data error
WARNING: ZLIB decompression error -3: data error
ERROR: Error uncompressing fp raster data for row 0 of <elevation>: error
       code -1
Segmentation fault (core dumped)

The last commit should fix this. We should test it more.