[FEATURE REQUEST] `geru` Triton kernel implementation

[LoserCheems]

Problem statement

The BLAS level-2 geru kernel (general rank-1 update) has no Triton implementation in kernel_course.triton_ops. The README BLAS table shows no Triton support for geru, leaving a classic BLAS-2 operation without a Triton example in this project.

Without a Triton geru kernel:

• users cannot study how rank-1 updates are mapped onto Triton’s execution model,
• there is no Triton baseline for performance comparison against PyTorch and CuTe geru implementations,
• the Triton backend is missing a key building block for more complex kernels.

Proposed solution

Add a Triton implementation of geru under kernel_course.triton_ops that matches the Python/PyTorch semantics and showcases tiling and memory access patterns for rank-1 updates.

Concretely:

• Introduce kernel_course/triton_ops/geru.py defining a Triton JIT kernel and Python wrapper.
• Implement $A = A + \alpha x y^\top$ by updating tiles of A using appropriate blocking strategies.
• Ensure numerical equivalence with the Python reference across supported dtypes.

Alternatives considered

Relying solely on PyTorch or CuTe for geru would:

• reduce opportunities to demonstrate rank-1 updates in Triton,
• leave the Triton column incomplete in the README BLAS table,
• limit the breadth of Triton examples in kernel-course.

Implementation details

• Add kernel_course/triton_ops/geru.py containing:
  • a Triton kernel operating on tiles of A with elements from x and y,
  • a wrapper that configures the grid/block sizes consistent with other Triton kernels.
• Handle arbitrary matrix sizes with guard checks.
• Integrate with future tests and benchmarks.

Use case

The Triton geru kernel will:

• illustrate a rank-1 update in Triton for educational purposes,
• provide a basis for performance comparisons to other backends,
• serve as a foundational operation for building more advanced kernels.

Related work

• Existing Triton kernels: triton_ops.copy, triton_ops.swap.
• Triton examples for matrix updates and BLAS-2 operations.

Additional context

This issue is part of expanding Triton coverage to BLAS level-2 operations listed in the README, including geru.