[quant] Add backend_independent option for quantized linear module (#48192)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/48192
This is to allow producing a backend independent quantized module
since some backend don't have packed weight for linear
Test Plan:
test_quantized_module.py
Imported from OSS
Reviewed By: raghuramank100
Differential Revision: D25061645
fbshipit-source-id: a65535e53f35af4f2926af0ee330fdaae6dae996
