Transfer resize iterator buffers without cloning

[bghira]

Summary

• move each completed resize buffer directly into its NumPy ndarray
• replace indexed storage with a queue so iteration can transfer ownership in O(1)
• make len(iterator) report the number of remaining results
• add coverage for values, shapes, exhaustion, and remaining length

Root cause

ResizeIterator.__next__ cloned the full Vec<u8> before constructing every ndarray. The API was described as zero-copy, but consuming it performed one avoidable full-output memory copy per image.

Performance

Median end-to-end iterator consumption on Apple M3 Max:

• 16 images, 1024x1024 -> 512x512: 2.831 ms -> 1.749 ms (38% faster)
• 64 images, 512x512 -> 224x224: 2.342 ms -> 1.646 ms (30% faster)

Validation

• added iterator ownership/length/exhaustion regression coverage
• cargo test --features opencv,simd (60 passed)
• Python test suite (69 passed)
• repository pre-commit hooks, including competitive performance benchmarks

[Copilot]

Pull request overview

This PR optimizes the OpenCV-backed Python resize iterator bindings to avoid per-item buffer cloning by transferring ownership of each completed resize buffer directly into the yielded NumPy ndarray, while also making len(iterator) reflect the number of remaining results.

Changes:

• Replace indexed buffer storage with a VecDeque so iteration can pop_front() and move buffers into ndarray without cloning.
• Update iterator construction paths to initialize/convert buffer storage appropriately.
• Add Python regression coverage for yielded values/shapes, exhaustion behavior, and remaining length reporting.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/python_bindings.rs	Switch ResizeIterator to queue-based ownership transfer and remove cloning during __next__.
tests/test_python_bindings.py	Add coverage asserting correct shapes/values, exhaustion, and len(iterator) remaining-count semantics.

---

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