XNNPACK quantized is failing for ic3 and ic4 when strict=False

[mergennachin]

#12368

cc @digantdesai @mcr229 @cbilgin

[mergennachin]

Here's an example: https://github.com/pytorch/executorch/actions/runs/16206315044/job/45757309536

[digantdesai]

Here's an example: pytorch/executorch/actions/runs/16206315044/job/45757309536

Core dump - fun!

[mcr229]

strange this is just a result of strict export?

[mergennachin]

Yeah with strict=True it is okay, but with strict=False, ic3/ic4 were failing -- it is succeeding with export, but during runtime it is throwing