RFC: Drop ppc64le

**Headline**: if you are a ppc user of conda-forge, please make yourself known with a comment below, describe some of the packages you use and how/where!

I propose that conda-forge stop supporting ppc64le.

In general, ppc seems to have dropped off significantly in HPC, to the point where it is unclear if conda-forge has more than 0 active ppc users.

ppc64le support has been very costly, both in maintainer time and CI costs, while also being the least widely used architecture across conda-forge.
As an increasingly rare architecture, it has often had problems seen on no other platforms, and cost a lot of time to debug.

It is hard to prove that users don't exist, but so far, every known ppc user who has been pinged about ppc support issues has said something to the effect of "I don't have use ppc anymore." To my knowledge, not one _current_ ppc user has spoken up in various conversations about ppc support troubles over the last couple of years.

## Supporting evidence

- > Summit has now been retired. It has been replaced by the amd64 machine Frontier. I'm not aware of any active ppc64le systems in US national HPC centers.
>
[ref](https://github.com/conda-forge/conda-forge.github.io/issues/2510#issuecomment-3382248736)

- > From a UK HPC colleague:
> > From my colleagues at the Cambridge HPC system:
> > > PowerPC is pretty uncommon these days. The only national (well, non-local) HPC system in the UK I’m aware of that still has it is [BEDE](https://bede-documentation.readthedocs.io/en/latest/index.html).
>
[ref](https://conda-forge.zulipchat.com/#narrow/channel/457337-general/topic/Cross-builds.20on.20linux-ppc64le.20are.20failing/near/570165358)
- > As an ex-user of HAL & ex-VAP in Statistics @ UIUC, I can attest that HAL was decommissioned on October 1st, 2025. The main reason was the system age and better systems came online for UIUC's campus, c.f. DeltaAI
>
[ref](https://conda-forge.zulipchat.com/#narrow/channel/457337-general/topic/Cross-builds.20on.20linux-ppc64le.20are.20failing/near/566480492)

- [CUDA dropped ppc support in 12.4](https://docs.nvidia.com/cuda/archive/12.4.0/cuda-toolkit-release-notes/index.html#deprecated-architectures)

- [Conda-Forge download statistics](https://github.com/conda-forge/conda-forge.github.io/issues/2510#issuecomment-2841566450) are consistent with ~100% of ppc downloads being from conda-forge's own CI (inconclusive)

## Options

1. stop _adding_ ppc64le builds, except by request (least disruptive), keep current low bar for manual removal
2. actively _remove_ ppc64le builds, and keep only those specifically requested (saves more maintainer and CI time, but needs to keep all infrastructure)
3. fully _prevent_ ppc64le builds, removing support from conda-smithy, conda-forge-pinning, etc. (most drastic because I'm not sure maintainers _can_ keep ppc builds if we do this).

I think we should do option 3 2.

## Tasks

- [ ] announce RFC, call for ppc users to speak up (I don't think any have in other conversations, yet)
- [ ] announce decision
- [ ] remove ppc from aarch migrator to stop getting new builds: https://github.com/regro/cf-scripts/pull/5862
- [ ] linter hint to suggest ppc removal: https://github.com/conda-forge/conda-smithy/pull/2504
- [ ] stop building _native_ ppc compiler packages? (not sure about this one)

### Originally suggested, but not necessary

- [ ] remove ppc from conda-forge-pinning
- [ ] remove ppc from docker-images
- [ ] remove ppc from conda-smithy
- [ ] removal migrator (only if remove on rerender is not going to be enough)

References:

- related discussion: https://github.com/conda-forge/conda-forge.github.io/issues/2510
- previous rfc, dropping pypy: https://github.com/conda-forge/conda-forge.github.io/issues/2255