Unify `par_dispatch`, `par_for_outer` & `par_for_inner` overloads #1269

Job	Run time
perf-and-regression (cuda, serial)	27m 31s
perf-and-regression (cuda, mpi)	1h 19m 44s
perf-and-regression-amdgpu (serial)	14m 37s
perf-and-regression (host, serial)	1h 11m 51s
perf-and-regression-amdgpu (mpi)	18m 39s
perf-and-regression (host, mpi)	32m 5s
	4h 4m 27s

Provide feedback