Adapt profiling script to handle multiple backends

[IgorSwat]

Problem description

As it stands for now, the profiling script (see GitLab) supports only XNNPACK-compiled models. However, with a little modification to the build script, it should be possible to allow handling other backends, most notably CoreML and Qualcomm.

What should be done

1. Modify the build_et.sh script to build ExecuTorch in a way to support other backends (CoreML, Qualcomm).

Benefits to React Native ExecuTorch

Handling this issue would make it easy to profile and debug models exported to backends such as CoreML or Qualcomm, preparing the ground for future tasks such as #556.

This PR might be helpful when merged: pytorch/executorch#19252

[msluszniak]

For the record, CoreML is supported by our library right now. However, we have no way to check which operations were delegated on CoreML backend which will be extremally helpful.