agents.md: improve effectiveness and add constraints

[GunniBusch]

---

• Have you followed the guidelines for contributing?
• Have you ensured that your commits follow the commit style guide?
• Have you checked that there aren't other open pull requests for the same formula update/change?
• Have you built your formula locally with HOMEBREW_NO_INSTALL_FROM_API=1 brew install --build-from-source <formula>?
• Is your test running fine brew test <formula>?
• Does your build pass brew audit --strict <formula> (after doing HOMEBREW_NO_INSTALL_FROM_API=1 brew install --build-from-source <formula>)? If this is a new formula, does it pass brew audit --new <formula>?

---

• AI was used to generate or assist with generating this PR. Please specify below how you used AI to help you, and what steps you have taken to manually verify the changes.

---

Made some improvements to agents.md, as I noticed some abnormalities in some recent PRs.

In my opinion, agents.md was not bad, but I think it was more a human quick guide than an LLM instruction, and not even a bad quick guide tbh.
But in my experience, the agents file, at least in these cases, should be less a direct instruction on basically how to run commands and do stuff, an more in giving general constraints.
Agents will fetch the docs anyway when instructed, so giving more general commands but finer edge-specific advice gave better output in my tests.

Some new things I added was an instruction for 1 commit/file/ pr (and instruction to squash) and one that maybe reduces the amount of PRs getting closed and new ones opened (but I do not really know if this is because of AI or ahm user😈).