fix quantized ic3/ic4/edsr examples (#574)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/574

The fix for quantized ic3/ic4/edsr requires runnign the pass

`convert_scalars_to_attrs` after capturing_pre_autograd_graph and before preparing and converting

We add using this pass in the examples to fix the quantized ic3/ic4/edsr

Reviewed By: digantdesai, guangy10

Differential Revision: D49850155

fbshipit-source-id: 36dca32611a53e27582c7a34eea178a65dc384f7