Fix Bug: Port over optimizations and fixes

[GillisC]

Description

• Corrections to memory alignment, single abstraction for making aligned memory allocations.
• Bug fix to Blocked GEMM
• Ported over corrected AVX2 and AVX512
• Add memory alignment allocation in Base64.hpp (there is probably a better way to do this.)
• Small fix to transpose which led to better performance when dealing with explicitly 2D tensors.

Issue

Closes #219
Closes #222

[Copilot]

Pull Request Overview

This PR applies bug fixes and optimizations across multiple GEMM implementations while standardizing memory alignment.

• Refactored memory alignment allocation and added runtime alignment checks.
• Improved tensor gemm implementations for blocked, AVX2, and AVX512 including shape validations and error message adjustments.
• Updated CMake configurations and utility modules to support the new alignment and optimized settings.

Reviewed Changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/openblas/CMakeLists.txt	Removed hard-coded memory alignment setting to rely on global config.
src/default/datastructures/tensor_gemm.cpp	Minor refactoring with removal of extraneous blank line.
src/default/datastructures/tensor.cpp	Added new get_raw_data() method.
src/default/datastructures/mml_array.cpp	Integrated aligned memory allocation and added alignment validation.
src/blocked/datastructures/tensor_gemm.cpp	Revised loop structure for better performance in blocked GEMM.
src/blocked/CMakeLists.txt	Removed legacy memory alignment definition.
src/avx512/datastructures/tensor_gemm.cpp	Enhanced shape validation and updated AVX512 gemm with corrected error messages.
src/avx512/CMakeLists.txt	Updated compile options using target_compile_options for AVX512.
src/avx/datastructures/tensor_gemm.cpp	Improved AVX2 gemm with shape checks and aligned memory usage; updated error messaging.
src/avx/CMakeLists.txt	Updated compile options for AVX2 with additional flags.
include/utility/base64.hpp	Modified memory allocation to support aligned tensor data.
include/utility/avx_mask_helper.hpp	Added helper functions for mask creation in AVX2/AVX512 GEMM.
include/utility/aligned_alloc.hpp	Introduced aligned memory allocation utility.
include/datastructures/tensor.hpp	Documented and declared the new get_raw_data() method.
include/datastructures/mml_array.hpp	Updated include dependencies for aligned allocation.
CMakeLists.txt	Revised global tensor alignment configuration with new options; potential duplicate condition identified.

[willayy]

A lot of code and im not really up to speed XD, but if it works its good :D

[GillisC]

@ehmc123 Quick question, if you look at avx_mask_helper that is currently sitting in the default folder, would i just be able to split it up and place the helper functions wherever in the avx and avx512 folders or is there some additional configuration that is needed in cmake?