Optimize bicubic resize with OpenMP + SIMD for large images

Summary

PR #1056 implements float-domain bicubic resize for Gemma4 HF parity. Performance is competitive for typical VLM images (≤2MP) but 5× slower than torchvision for 12MP+ images.

Current performance (resize only, no JPEG decode)

Why torchvision is faster for large images

Torchvision uses ATen's _upsample_bicubic2d_aa kernel (aten/src/ATen/native/cpu/UpSampleKernel.cpp) which has:

• Multi-threaded execution via at::parallel_for (OpenMP)
• AVX2/NEON SIMD vectorization (specialized uint8 paths, 8 floats per instruction)
• Precomputed weight tensors for batched multiply-add
• ~16ms fixed overhead from tensor creation and thread pool setup (why it's slower for small images)

Proposed optimizations

1. OpenMP parallelism for vertical pass: The vertical resample loop iterates over output rows independently — adding #pragma omp parallel for would parallelize across cores with minimal code change.

2. AVX2 intrinsics for horizontal kernel: The horizontal inner loop does 3 multiply-adds per kernel tap — could process 8 taps simultaneously with _mm256_fmadd_ps.

3. Row-batch vertical processing: Process multiple output columns per iteration to improve instruction-level parallelism in the vertical pass.

Notes

• No precedent for OpenMP or SIMD in ort-extensions' own source code (checked: shared/, operators/). All existing image processing is single-threaded scalar C++.
• Would need CMake changes to enable OpenMP and SIMD compiler flags.
• The current implementation uses the same algorithm as Pillow/torchvision (Keys cubic a=-0.5, separable, antialias) and matches torchvision output to within 5×10⁻⁵ max pixel diff.

Priority

Low — current performance is acceptable for typical VLM images (≤2MP). Only impacts 12MP+ unresized phone photos, which are uncommon in production VLM pipelines.

References

• PR #1056: Float-domain bicubic resize for HF parity
• PyTorch source: aten/src/ATen/native/cpu/UpSampleKernel.cpp (lines 1994–2021)

Start by reading PR #1056 and the current bicubic resize implementation in the repository's shared/ and operators/ areas, then compare the approach with PyTorch's UpSampleKernel.cpp. Benchmark the existing large-image path before evaluating OpenMP, SIMD, and row-batch changes. Done means improved performance for 12MP+ images while preserving the stated maximum pixel difference and keeping smaller-image performance acceptable.