Added faster version of matrix vector multiply that is used when not … #2975

Job	Run time
codecheck / codecheck	44m 13s
conda / conda (ubuntu-20.04)	11m 2s
docker / docker (fedora39)	20m 2s
conda / conda (macos-12)	23m 27s
docker / docker (rocky8)	22m 43s
linux / linux (-doc-mpi-)	38m 37s
linux / linux (-mpi-)	49m 24s
conda / conda (macos-14)	7m 41s
mac / macports (macos-12)	17m 54s
linux / linux (-coverage-mpi-)	1h 0m 27s
mac / macports (macos-14)	9m 40s
mac / macports (+allmodules, macos-12)	21m 54s
mac / macsimple (macos-12)	8m 35s
linux / linux (-debug-)	30m 26s
mac / macsimple (macos-14)	6m 1s
linux / linux (-debug-mpi-)	40m 43s
linux / linux (-intel-)	43m 49s
linux / linux (-pycv-mpi-)	16m 47s
	7h 53m 25s

Provide feedback