Qualcomm AI Engine Direct - Move soc infomation from cpp to python (#1679)

Summary:
- Using flatbuffer as compile spec to communicate between python and cpp
 - Collect soc information into backends/qualcomm/serialization/qnn_compile_spec_schema.py which is convenient to add more SoC
 - Add flatbuffer into cmake build flow
 - Add tutorial to add new soc supported
 - Set SoCModel instead of arch because arch will be deprecated by QNN. (Generating deprecated warning after QNN 2.14)
 - Add a macro wrapper for log with different log level

Pull Request resolved: https://github.com/pytorch/executorch/pull/1679

Reviewed By: JacobSzwejbka

Differential Revision: D53191783

Pulled By: cccclai

fbshipit-source-id: 3c2a7a7633480be4541f542ed339cb5e491438df