mpi: fix in-loop allreduce with multiple reductions #8265

Job	Run time
pytest-ubuntu-py312-gcc12-cxxomp-base	55m 47s
pytest-ubuntu-py310-gcc14-omp-base	45m 12s
pytest-ubuntu-py312-gcc13-omp-adjoint	36m 44s
pytest-ubuntu-py311-gcc11-cxxnoomp-base	45m 38s
pytest-ubuntu-py312-gcc12-cxxomp-adjoint	40m 7s
pytest-ubuntu-py310-gcc14-omp-adjoint	32m 28s
pytest-ubuntu-py312-gcc13-omp-base	51m 50s
pytest-ubuntu-py311-gcc11-cxxnoomp-adjoint	33m 1s
pytest-ubuntu-py311-gcc10-noomp-base	42m 29s
pytest-ubuntu-py310-gcc9-omp-base	44m 25s
pytest-ubuntu-py310-gcc9-omp-adjoint	32m 51s
pytest-docker-py310-gcc-omp-base	37m 24s
pytest-ubuntu-py311-gcc10-noomp-adjoint	30m 46s
pytest-osx-py312-clang-omp-base	33m 6s
pytest-ubuntu-py313-gcc14-omp-base	53m 45s
pytest-ubuntu-py313-gcc14-omp-adjoint	39m 22s
pytest-docker-py310-icx-omp-base	31m 58s
pytest-docker-py310-icx-omp-adjoint	28m 11s
pytest-docker-py310-gcc-omp-adjoint	28m 16s
	12h 23m 20s

Provide feedback