Arm backend: Add Arm Neural Statistics

[wwwind]

This PR adds opt-in VGF Arm Neural Accelerator Statistics collection for ETDump.

When EXECUTORCH_VGF_ENABLE_NEURAL_STATISTICS=1 is set, the VGF backend collects VK_ARM_data_graph neural statistics metadata, wraps the opaque binary blobs in a versioned JSON payload, and attaches it to ETDump delegate metadata under the VGF_NEURAL_STATISTICS event. The implementation keeps normal timing-only profiling unchanged by default, handles unsupported Vulkan/API/bind-point paths gracefully, and adds Inspector parsing support to recover decoded debug_database, statistics_info, and statistics_memory bytes.
This option EXECUTORCH_VGF_ENABLE_NEURAL_STATISTICS was added to attach this payload only when it is requested specifically, otherwise it could blow up ETDump in some runs.

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20662

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit d57dd13 with merge base d03dee5 ():

NEW FAILURE - The following job has failed:

• pull / test-qnn-python-imports-linux / linux-job (gh)
  RuntimeError: Command docker exec -t 7d7f2f1e8faac844afb58dbd8150939129a7d3b6f18d272628cb656b6a15946a /exec failed with exit code 92

This comment was automatically generated by Dr. CI and updates every 15 minutes.