Support per channel tensor with unpacking in QNNPACK (#96268)
Summary: Supporting Per Channel quantizer with unpacking for QNNPACK.
Test Plan: buck2 test //caffe2/test/quantization:quantization -- test_qlinear_per_channel_qnnpack_free_memory_and_unpack
Differential Revision: D43865268
Pull Request resolved: https://github.com/pytorch/pytorch/pull/96268
Approved by: https://github.com/kimishpatel