Add xnnpack global average pool op (#55791)
Summary:
Adaptive average pool with output size (1, 1) is a global average pool
For mobile use xnnpack to speed up that path
Pull Request resolved: https://github.com/pytorch/pytorch/pull/55791
Test Plan:
buck test //xplat/caffe2:pt_xnnpack_test
pytest test/test_xnnpack_integration.py::TestXNNPACKOps
Fixes #{issue number}
Reviewed By: kimishpatel
Differential Revision: D27711082
Pulled By: axitkhurana
fbshipit-source-id: 8757042c4a31a60451d8ba5fb6bf8cfbaf0a8d10
