Support CUDA 12.6.

[paulstansifer]

I updated build.rs and Cargo.toml and ran the bindgen script.

Local testing mostly passed; all of the errors are the same, which is probably indicating that I still haven't set everything up correctly on my machine.

---- driver::safe::launch::tests::test_launch_with_views stdout ----
thread 'driver::safe::launch::tests::test_launch_with_views' panicked at src/lib.rs:98:5:
Unable to dynamically load the "cuda" shared library - searched for library names: ["cuda", "nvcuda"]. Ensure that `LD_LIBRARY_PATH` has the correct path to the installed library. If the shared library is present on the system under a different name than one of those listed above, please open a GitHub issue.

---- driver::safe::launch::tests::test_par_launch stdout ----
thread 'driver::safe::launch::tests::test_par_launch' panicked at src/lib.rs:98:5:
Unable to dynamically load the "cuda" shared library - searched for library names: ["cuda", "nvcuda"]. Ensure that `LD_LIBRARY_PATH` has the correct path to the installed library. If the shared library is present on the system under a different name than one of those listed above, please open a GitHub issue.

failures:
    cublas::safe::tests::test_dgemm
    cublas::safe::tests::test_dgemv
    cublas::safe::tests::test_sgemm
    cublas::safe::tests::test_sgemv
    cublaslt::safe::tests::test_matmul_f32
    cudnn::safe::tests::test_conv1d
    cudnn::safe::tests::test_conv2d_pick_algorithms
    cudnn::safe::tests::test_conv3d
    cudnn::safe::tests::test_create_descriptors
    cudnn::safe::tests::test_reduction
    curand::safe::tests::test_different_seeds_neq
    curand::safe::tests::test_log_normal_f32
    curand::safe::tests::test_log_normal_f64
    curand::safe::tests::test_normal_f32
    curand::safe::tests::test_normal_f64
    curand::safe::tests::test_rc_counts
    curand::safe::tests::test_seed_reproducible
    curand::safe::tests::test_set_offset
    curand::safe::tests::test_uniform_f32
    curand::safe::tests::test_uniform_f64
    curand::safe::tests::test_uniform_u32
    driver::safe::alloc::tests::test_copy_uses_correct_context
    driver::safe::alloc::tests::test_device_copy_to_views
    driver::safe::alloc::tests::test_leak_and_upgrade
    driver::safe::alloc::tests::test_post_alloc_arc_counts
    driver::safe::alloc::tests::test_post_build_arc_count
    driver::safe::alloc::tests::test_post_clone_arc_slice_counts
    driver::safe::alloc::tests::test_post_clone_counts
    driver::safe::alloc::tests::test_post_release_counts
    driver::safe::alloc::tests::test_post_take_arc_counts
    driver::safe::alloc::tests::test_slice_is_freed_with_correct_context
    driver::safe::core::tests::test_transmutes
    driver::safe::launch::tests::test_large_launches
    driver::safe::launch::tests::test_launch_with_16bit
    driver::safe::launch::tests::test_launch_with_32bit
    driver::safe::launch::tests::test_launch_with_64bit
    driver::safe::launch::tests::test_launch_with_8bit
    driver::safe::launch::tests::test_launch_with_floats
    driver::safe::launch::tests::test_launch_with_mut_and_ref_cudarc
    driver::safe::launch::tests::test_launch_with_views
    driver::safe::launch::tests::test_mut_into_kernel_param_no_inc_rc
    driver::safe::launch::tests::test_par_launch
    driver::safe::launch::tests::test_ref_into_kernel_param_inc_rc
    driver::safe::threading::tests::test_threading
    nccl::result::tests::multi_thread
    nccl::result::tests::single_thread
    nccl::safe::tests::test_all_reduce

test result: FAILED. 137 passed; 47 failed; 1 ignored; 0 measured; 0 filtered out; finished in 0.36s.

[paulstansifer]

This should fix #280

[coreylowman]

This looks good (thanks for the contribution!). Looks like the NCCL/CUDNN 12060 versions weren't included (seems like the 12.5 files were modified). Seems like yu should be able to generate using this docker image: docker pull nvidia/cuda:12.6.0-cudnn-devel-ubuntu22.04

[carlos-verdes]

@paulstansifer it looks like there are couple of format issues to be fixed in order to merge this PR, are you working on it?

[paulstansifer]

I'm currently the parent on parental leave, so my availability is irregular, but I'll probably be able to get back to it in a couple days.

[carlos-verdes]

I have two kids, second is 5 months so I feel you, take your time

[paulstansifer]

I've updated cudnn and nccl. Unfortunately, I think that doing so obliterated the other failing checks that GitHub was reporting. I can see cargo clippy is complaining, but it's about files that this PR doesn't touch, and I tried running cargo fmt, but it didn't seem to modify anything. Should I just roll fixes for the clippy complaints into this? Seems like it's a few minor formatting things.

[coreylowman]

This looks good now! Thanks for the update & tagging me with the re-review