fix(detector): detect direct self matmul output

[prasannakotyal]

Summary

• extend SELF_MATMUL_OUTPUT to direct self-matmul calls through torch.mm/torch.matmul aliases
• detect data @ data.t() and equivalent mm(data, data.t()) entrypoint returns, including import torch as T aliases
• preserve the existing narrow functools.partial(helper) self-matmul detection

Target

KernelGuard-Red-Submission: 283

Validation

• UV_CACHE_DIR=/tmp/uvcache uv run python -m py_compile kernelguard.py
• import torch as T; return T.mm(data, data.t()): classification=hacked, should_filter=true, pattern SELF_MATMUL_OUTPUT
• from torch import mm as matmul; return matmul(data, data.t()): classification=hacked, should_filter=true, pattern SELF_MATMUL_OUTPUT
• UV_CACHE_DIR=/tmp/uvcache uv run python ../../kernelguard_bypasses/eval_blue_patch.py kernelguard.py clean fixtures remain should_filter=False
• Partial self-matmul detector passed official eval on PR fix(detector): detect partial self matmul output #254 with TP 20/20, FP 20/20, surgicalness 1.0

[github-actions]

KernelGuard Blue Evaluation

• Evaluation job: 200
• Blue submission: 172
• Result: passed
• Target red submission caught: True
• Validation suite: TP 20/20, FP 20/20
• Surgicalness score: 1.0
• Workflow run: https://github.com/SinatrasC/kernelguard/actions/runs/25383868951

[SinatrasC]

Thanks for the KernelGuard Flywheel Campaign contribution. This PR is now superseded by the consolidated rule-family implementation in #273, which folds this detector coverage together with the related passing-eval variants.