Enable even more unit tests for warp size 32

[sstamenk]

Enables more tests for warp size 32 Radeon GPUs for #1748. Out of the enabled tests 60 pass and only the following test is failing due to accuracy issues:

• tests/test_functional.py::TestIGEMMFunctional::test_minmax_igemm[transpose=F-batch_dim=2-hidden_dim=4096-seq_dim=32]

Attaching logs bitsandbytes_even_more_tests_gfx1201.log

[sstamenk]

@electron271 fixed the linting issues for the upstream PR as well, please take a look when you get the chance to

[electron271]

@electron271 fixed the linting issues for the upstream PR as well, please take a look when you get the chance to

the upstream PR was already merged so this would need to be another PR to bitsandbytes

[electron271]

(misclick)

[sstamenk]

Closing this and opening a new PR to upstream, thanks.