[Performance] consolidate TDs in ParallelEnv without buffers #10505

Job	Run time
build-wheel-linux (3.8, cp38-cp38, --index-url https://download.pytorch.org/whl/cpu, "['cpu', '11...	1m 37s
build-wheel-windows (3.8, 3.8)	2m 21s
build-wheel-linux (3.9, cp39-cp39, --index-url https://download.pytorch.org/whl/cpu, "['cpu', '11...	1m 38s
build-wheel-windows (3.9, 3.9)	2m 51s
build-wheel-linux (3.10, cp310-cp310, --index-url https://download.pytorch.org/whl/cpu, "['cpu', ...	1m 45s
build-wheel-windows (3.10, 3.10.3)	3m 27s
build-wheel-linux (3.11, cp311-cp311, --index-url https://download.pytorch.org/whl/cpu, "['cpu', ...	1m 37s
build-wheel-windows (3.11, 3.11)	2m 57s
test-wheel (linux, ubuntu-20.04, 3.8)	2m 52s
test-wheel (linux, ubuntu-20.04, 3.9)	2m 14s
test-wheel (linux, ubuntu-20.04, 3.10)	2m 48s
test-wheel (linux, ubuntu-20.04, 3.11)	2m 58s
test-wheel-windows (3.8)	5m 19s
test-wheel-windows (3.9)	3m 35s
test-wheel-windows (3.10)	3m 45s
test-wheel-windows (3.11)	3m 54s
	45m 38s

Provide feedback