Revert ExecuTorch pack to 1.1.0 until Fusion Studio supports 1.3.1

[MatthiasHertelArm]

Reverts #14 (21acbaf, "Bump ExecuTorch pack to 1.3.1-rc8").

Why

The Fusion Studio worker services still compile with ExecuTorch 1.1. ModelNova workspaces uploaded to Fusion Studio for local compilation currently fail to build because of:

• an ExecuTorch version mismatch in ai_layer.clayer.yml (the 1.3.1-rc8 pack does not exist on the workers), and
• the changed ML component identifiers introduced with the 1.3.x pack layout (operators moved to the ExecuTorch Operators Cgroup; Platform Bare-Metal / Stubs RandomOps dropped).

This restores the vendored PyTorch::ExecuTorch 1.1.0-rc1-build.12 pack, the 1.1.0 component selection, and the cpackget lines in the CI workflows until the webinar is over and Fusion Studio has been updated to ExecuTorch 1.3.1.

Notes

• ai_layer.clayer.yml moved to algorithm/ML/ai_layer/ in Integrate Common Inference Library, Optimize Memory Usage, and Enhance Display Configuration #13 after the bump, so the revert was resolved at the new location (1.1.0 pack path at the new relative depth, 1.1.0 component set). The restored 1.1.0 pack content is byte-identical to the pre-Bump ExecuTorch pack to 1.3.1-rc8 #14 state.
• The ExecuTorch 1.3.1 state of main is preserved on the executorch-1.3.1 branch for re-landing after the webinar (alternatively, revert this revert).
• Integrate Common Inference Library, Optimize Memory Usage, and Enhance Display Configuration #13 (CIL integration) and Remove unused files and add default comments to AppKit-E8 RTE_Device.h #16 are unaffected: the runner uses the same ExecuTorch API surface as before the bump, and model_pte.h keeps the same .pte format identifier, so it loads on the 1.1 runtime.