roll-out of glibc 2.39 / RHEL10-derived infrastructure

This is a continuation of #1941 and #2342. After the end of CentOS (#1436), we've were forced to choose between some various ABI-compatible RHEL clones; the most prominent of those are Almalinux and Rockylinux, and at the time, we ended up going with Alma (though I don't recall any particularly pressing reason for that choice over Rocky, though they were a step further behind with ppc64le bring-up IIRC).

This has worked out well so far, but in the context of #1744, we're running into the issue that alma does not support riscv64[^1], i.e. we cannot use their RPMs for this. As it turns out, rocky linux [added](https://rockylinux.org/news/rockylinux-support-for-riscv) riscv support for rocky v10. Since all three distros variants _should_ in principle be ABI-compatible with each other, I don't think this would be such a big hurdle.

It does of course create more work than if things had just been available on alma, but it'd be even more work to pull our infra from two different distros (with different package metadata, and occasional packaging differences), so IMO it'd make sense to build our RHEL10-based infra on Rocky

[^1]: check content of https://repo.almalinux.org/almalinux/10.0/BaseOS, for example. I found an old post on their [forums](https://forums.almalinux.org/t/add-support-for-riscv64/3402), but it has 0 responses.

* [x] https://github.com/conda-forge/docker-images/pull/319
* [x] https://github.com/conda-forge/linux-sysroot-feedstock/pull/92
* https://github.com/conda-forge/linux-sysroot-feedstock/pull/94
* https://github.com/conda/conda-build/issues/5853
* [ ] update https://github.com/conda-forge/cdt-builds
* [x] https://github.com/conda-forge/conda-forge-pinning-feedstock/pull/8727

If we switch to rocky:
* https://github.com/conda-forge/linux-sysroot-feedstock/pull/93
* build new rocky-based images
* make sure CDTs are from rocky too

There's also a renewed [discussion](https://github.com/conda-forge/cdt-builds/issues/89) if we can further reduce the set of CDTs, in addition to what we did going from cos7 to alma8. I'm sympathetic to that goal, but we need to manage the security side of this, which is (IMO) non-trivial.