Fix Llama transform heap overflow for non-square tile size

[nenad1002]

No description provided.

[Copilot]

Pull request overview

This PR hardens the Llama3ImageTransform image pre-processing path to prevent a heap out-of-bounds write when given a non-square tile size configuration, and adds a regression test/fixture to ensure invalid configs are rejected.

Changes:

• Add validation in Llama3ImageTransform::Init() to reject non-square (and non-positive) tile sizes.
• Add a defense-in-depth bounds check in DoPad() to prevent overrunning the padded image allocation.
• Add a regression test and JSON fixture that attempts to create a processor with a non-square tile size.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
test/pp_api_test/test_processor.cc	Adds a regression test ensuring non-square tile configs are rejected.
test/data/processor/mllama/llama_3_image_nonsquare.json	Adds a failing processor configuration fixture with a non-square "size".
shared/api/image_transforms_mllama.hpp	Adds safety checks in DoPad() and rejects non-square tile sizes in Init().

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.