Add support for batching to AutoDiffCompostion in PyTorch mode. #10159

Job	Run time
build (3.8, x64, ubuntu)	1h 37m 21s
build (3.8, x64, windows)	1h 31m 41s
build (3.11, x64, ubuntu)	1h 17m 46s
build (3.11, x64, macos)	36m 43s
build (3.11, x64, windows)	1h 36m 4s
build (3.12, x64, ubuntu)	1h 18m 16s
build (3.12, x64, macos)	41m 28s
build (3.12, x64, windows)	1h 31m 1s
build (3.9, macos, --cov=psyneulink)	47m 35s
build (3.9, ubuntu, --forked -m "not llvm")	40m 40s
build (3.9, x86, windows)	1h 35m 17s
build (3.10, ubuntu, --fp-precision=fp32)	1h 17m 1s
build (3.10, macos, -m benchmark --benchmark-enable --benchmark-only --benchmark-min-rounds=2 --b...	28m 42s
build (3.8, macos, min)	44m 45s
	15h 44m 20s

Provide feedback