[mpich romio 5a036e7] mpl/cuda: Fix header include order

> Pulled in from mpich romio, branch "main".
> Their commit message is below.
>
> This is part of a batch of commits from the
> following set of PRs:
>  *  https://github.com/pmodels/mpich/pull/4943
>     -- darray fix which contains a flatten fix
>         73a3eba
>         c4b5762
>  *  https://github.com/pmodels/mpich/pull/4995
>     -- write strided fix
>         bbb5210
>  *  https://github.com/pmodels/mpich/pull/5100
>     -- build fix for -Wpedantic
>         ad0e435
>  *  https://github.com/pmodels/mpich/pull/5099
>     -- build fix, they had let file-system=...gpfs bit rot
>         e1d42af
>         313289a
>         83bbb82
>  *  https://github.com/pmodels/mpich/pull/5150
>     -- build fix, configure-related _GNU_SOURCE
>         a712b56
>         5a036e7
>  *  https://github.com/pmodels/mpich/pull/5184


>     -- build fix, continuation of _GNU_SOURCE fix
>         d97c4ee

mpl.h must be included before any system headers so that config
settings are taken into account. Fixes an issue where aligned_alloc
may be used without a declaration.

Signed-off-by: Mark Allen <markalle@us.ibm.com>

3rd-party/romio341/mpl/src/gpu/mpl_gpu_cuda.c

@@ -3,8 +3,8 @@
  *      See COPYRIGHT in top-level directory.
  */
 
-#include <dlfcn.h>
 #include "mpl.h"
+#include <dlfcn.h>
 #include <assert.h>
 
 #define CUDA_ERR_CHECK(ret) if (unlikely((ret) != cudaSuccess)) goto fn_fail