build

ACC: use data regions rather than hard-coded parallel gang loops #642

Job	Run time
ci (Release, DP, linux gnu-10)	35m 9s
ci (Release, DP, linux gnu-14)	45m 41s
ci (Release, DP, linux nvhpc-21.9)	32m 40s
ci (Release, DP, linux intel-classic)	16m 37s
ci (Release, DP, macos)	37m 8s
ci (Release, SP, linux gnu-10)	34m 36s
ci (Release, SP, linux gnu-14)	47m 0s
ci (Release, SP, linux nvhpc-21.9)	33m 26s
ci (Release, SP, linux intel-classic)	16m 7s
ci (Release, SP, macos)	37m 49s
	5h 36m 13s

Provide feedback