fix: AIX searches dynamic libraries in `LIBPATH`.
### What does this PR try to resolve?
On IBM AIX machines people have encountered issues in `compiletest` and rustc's bootstrap builder. They haven't encountered any in cargo. This PR is made for avoiding potential failures in the future in cargo.
It's documented in <https://www.ibm.com/support/pages/libpath-environment-variables-aix-platforms>:
> The `LIBPATH` environment variable tells AIX applications where to find shared libraries when located in a different directories than those specified in the header section of the executable.
See also the counterpart in <https://github.com/rust-lang/rust/pull/109526>
### How to verify and test this in Cargo's CI?
This is indeed an issue. At IBM people are maintaining a buildbot since GitHub Action doesn't support AIX yet.
ci: check if any version bump needed for member crates
### What does this PR try to resolve?
Check if a crate requires a version bump in CI.
Part of #12033
### How should we test and review this PR?
Demo action result: <https://github.com/weihanglo/cargo/actions/runs/4941952784/jobs/8835049007>
### Additional information
This doesn't devalue #12089. I love the changelog detection, saving maintainers' times a lot ( I am the one writing changelogs for each release for now). However, the amount of code there is too excessive to me. I think someday when we are going to integrate [cargo-release](https://crates.io/crates/cargo-release) and/or [ehuss/cargo-new-release](https://github.com/ehuss/cargo-new-release), we can remove this script perhaps entirely :)
I tried to document the script a bit hard. Hope it won't get worse over time.
These tree packages are considered to be published something.
To reduce the logic of xtask-unpubilshed, let's flag them false for now.
We can always set it `true` when needed.
feat: Add `-Zmsrv-policy` feature flag
### What does this PR try to resolve?
Nothing noticeable....
The intent is to unblock experiments with different compatible MSRV policies like
- #9930
- #10653
- #10903
While I normally don't like PRs that do nothing on their own, this at least allows any one of those efforts to move forward with different people without juggling these base commits for whoever is first to include in their PR
While there isn't an RFC for this yet, this is intended to allow us to experiment to get a better idea of what we should put in an RFC. In some cases, we first do an eRFC for this but I assumed this wouldn't be needed in this case as this builds on rust-lang/rfcs#2495 and, I'm assuming, will be more surgical in nature
### How should we test and review this PR?
The `Summary` changes are largely untested as they will be mostly tested through the future work that builds on this PR. However, I wasn't too concerned about that because the code is relatively trivial.
### Additional information
I chose the name `msrv-policy` to distinguish this unstable feature from `rust-version`. Though those appear in different places (`Cargo.toml` vs `-Z`), I can see them being confusing which was especially apparent when editing `unstable.md` which has an anchor for `rust-version`.
