[llvm] [docs] Strengthen our quality standards and connect AI contribution policy to it (PR #154441)

[Reid Kleckner via llvm-commits]

rnk wrote:

> Some high level feedback:
> 
> * I think this is a bit too verbose for a policy document. Some of this (like the references to other AI policies) are something to mention in an RFC to justify how we arrived here, but they are not something every person reading the finished policy needs to look at. Similarly, I'm not sure we should be including the meta philosophy around "extractive" contributions in the policy document. We should be able to explain this more concisely -- is the contribution worth the required investment of reviewer time?

My view is that these references are a good way of capturing the "alternatives considered", which is often the most important part of any technical design document. They are also likely to be helpful in capturing the "why" of any policy, which is something that the next policy doc editor will probably care about (I now find myself in this position). I've personally received a lot of links to what other projects are doing, and so the section serves as a way of saying "yes, we looked at all that, and decided we want to make compromise X".

I think the Developer Policy doc is just growing, and it needs to be split up (LangRef has grown similarly), and it also needs to be adapted to apply to the entire LLVM umbrella project. Much of the text is focused on LLVM the optimizer and code generator.

Re: extractive contributions, @chandlerc gave me the opposite feedback, that I should center the ideas of extractive and non-extractive contributions. I defer to the opinions of active maintainers, but that's why I went philosophical. Quality is inherently subjective and philosophical, at the end of the ady. 

> * One of the motivations behind revisiting this policy is to avoid wasting reviewer time on low-quality AI generated PRs. I think what I'm currently missing from this policy is a means by which a reviewer can reject a low-quality PR without significant justification. Basically, this needs to be at the level of rejecting a PR with a copy&paste comment pointing to the policy.

This is true, I will take action on that, I got the same feedback from Aaron. I do need to get that PR-copy-pasta message into the proposal near the top.

https://github.com/llvm/llvm-project/pull/154441