[FEATURE REQUEST] `geru` PyTorch kernel implementation

[LoserCheems]

Problem statement

The BLAS level-2 geru kernel (general rank-1 update) is not implemented for the PyTorch backend in kernel_course.pytorch_ops. The README BLAS table lists no PyTorch support for geru, leaving this important operation absent from the PyTorch kernel set.

Without a PyTorch geru kernel:

• there is no unified API for $A = A + \alpha x y^\top$ on PyTorch tensors within kernel-course,
• cross-backend comparisons for rank-1 updates are not possible,
• higher-level modules must use ad-hoc PyTorch code instead of a standard kernel abstraction.

Proposed solution

Add a PyTorch implementation of geru under kernel_course.pytorch_ops that matches the Python reference semantics and follows existing backend conventions.

Concretely:

• Introduce kernel_course/pytorch_ops/geru.py implementing $A = A + \alpha x y^\top$ for PyTorch tensors.
• Accept scalar alpha, vector x, vector y, and matrix A.
• Use idiomatic PyTorch operations (e.g. outer product plus in-place add) while maintaining correctness.

Alternatives considered

Relying exclusively on raw PyTorch operations outside of a dedicated kernel would:

• fragment API usage across the project,
• complicate benchmarking geru across backends,
• weaken the pedagogical mapping between mathematical kernels and backend implementations.

Implementation details

• Add kernel_course/pytorch_ops/geru.py with a public geru function.
• Validate shapes so that A has dimensions (m, n), x has length m, and y length n.
• Ensure consistent dtype and device handling with other PyTorch kernels.
• Update kernel_course/pytorch_ops/__init__.py to expose geru.

Use case

The PyTorch geru kernel will:

• provide a standard rank-1 update primitive for use in higher-level modules,
• serve as a target for cross-backend performance and correctness comparisons,
• help illustrate BLAS level-2 patterns in PyTorch.

Related work

• Existing PyTorch kernels: pytorch_ops.copy, pytorch_ops.swap.
• PyTorch outer product and rank-update patterns.

Additional context

This issue contributes to completing the geru row for the PyTorch column in the README BLAS table.