feat(meta): add agent-review rubric and example GitHub Action

[jfindlay]

• AGENTS-REVIEW.md: A vendor-neutral, rubric for a single review pass on a Physlib PR.
• .github/workflows/agent-review.yml: An example implementation triggered by the agent-review label.

[github-actions]

Thank you for this PR, which will now be reviewed.
If submitting to ./Physlib or ./QuantumInfo, please
see our review guidelines
if you are not familiar with the process. You should expect a back and forth
with a reviewer before your PR is merged. See also that link for how to
add appropriate labels to your PR. The PR will also go through a number
of automated checks. You can learn more about these here,
including how to run them locally.

If you are submitting to ./PhyslibAlpha there will be a lighter review process,
though your PR must still pass the automated checks.

If you want to bring attention to this PR, please write a message on this
thread of the Lean Zulip.

[doxtor6]

sonnet 4.5 is not good enough. I think we should use the top tier models here like opus-4.8 or gpt-5.5

[jstoobysmith]

Do we know who actually provides the agents for this? I assume we would need some AI key.

This might be a use case of #1211

[jfindlay]

sonnet 4.5 is not good enough. I think we should use the top tier models here like opus-4.8 or gpt-5.5

Sonnet 4.5 is a compromise on token cost. There are some other token optimizations in AGENTS-REVIEW.md as it's currently written, like only running the review on explicit (re)label. I agree that some aspects of the review would be better at a higher tier agent, but more mechanical tasks like assessing docs coverage could be given to a lower tier agent to save cost.

Do we know who actually provides the agents for this? I assume we would need some AI key.

Proprietary frontier agent models are less generous with their free tier than for example GitHub itself. I do not know much about this area, but since this is a high profile, important, public project, there could be a way to get a sponsorship or a discount or perhaps GitHub or some libre software consortium has a program for such.

[doxtor6]

Considering the current PR traffic of PhysLib, I think there is no need to compromise on token cost. I also generally do not trust models weaker than Opus-4.6; they sometimes produce serious hallucinations and require a lot of engineering effort to make them work properly.

I am still reading the GitHub agentic workflows documentation. It seems like it supports AI model subscription tiers, in which case a 100-200usd/month subscription should definitely be enough.