[torchao]: fix how FqnConfig is resolved and handled (#14686)

* fix: how fqnconfig passed to torchaoconfig is resolved.

* testing with quantize_()

* change torchao version in setup.py as well.