Implement lazy transpose for Vulkan GPU backend with stride swapping

Implement Vulkan GPU contiguous() method with gather compute shader

Add contiguous() method and improve non-contiguous tensor handling

Refactor idxunravel implementation and add comprehensive indexing tests

Refactor tensor indexing with TensorLayout class and standalone functions

Start index operations refactoring with strides

Add slicing to the TensorProperties class

Запрос на слияние 'matmul-transpose' (#62) из matmul-transpose в master

Update new sum method interface usages

Add optional bias to Linear layer, remove transpose in linear bwd and generalize sum reduction

Update new sum method interface usages

Add optional bias to Linear layer, remove transpose in linear bwd and generalize sum reduction

Update style, add example

update: articles.md

create: new file