Arm backend: Remove duplicate scratch allocation in Direct Drive

[Erik-Lundell]

Scratch was always allocated using the temporary allocator, even
though Direct Drive allocated it itself in invoke_linux_driver.

Introduce platform_allocate to let each platform decide for itself
how to allocate memory. Some extra handling is needed around
copying io, though this will be addressed properly in a later patch.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22175

• 📄 Preview Python docs built from this PR

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

❌ 3 New Failures, 1 Unrelated Failure

As of commit 0f280f1 with merge base 6ce2715 ():

NEW FAILURES - The following jobs have failed:

• pull / unittest-nxp-neutron / linux-job (gh)
  ../../dev/ir/converter/node_converter/test_conv_converter.py::TestConv::test__d_fwd_basic[QAT-basic
• trunk / test-huggingface-transformers-xnnpack (smollm2-135m|xnnpack|--quantize) / linux-job (gh)
  RuntimeError: Command docker exec -t eff7b0dbd30e37fa28c040aaa5367c757a86ed000cfe996e6300bb919e901941 /exec failed with exit code 1
• trunk / test-llama-runner-mac (fp32, coreml) / macos-job (gh)
  executorch

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• pull / test-qnn-testsuite-linux / test-backend-linux (qnn, models) / linux-job (gh) (trunk failure)

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

[rascani]

Hey @Erik-Lundell - just as an fyi, I've been working on a proof of concept for delegate scratch bufffers: #21985.

[Erik-Lundell]

Hey @Erik-Lundell - just as an fyi, I've been working on a proof of concept for delegate scratch bufffers: #21985.

That looks great @rascani. I will drop the second commit of the PR and wait for your solution.