Move cuPQC ML-KEM upstream away from personal repo

[praveksharma]

The cuPQC ML-KEM upstream added by #2044 resides in a personal GitHub repo. Either move this upstream to a Nvidia repo or update copy_from_upstream.py to accomodate a new class of local upstream.

[baentsch]

move this upstream to a Nvidia repo

That's my clear preference: It makes clear whose code this is and who has the responsibility (and capability) to update the code with future updates to the NVIDIA library or any potential breakages in cuPQC becoming visible only in liboqs or downstream.

or update copy_from_upstream.py to accomodate a new class of local upstream.

What do you understand by "local upstream"? Either something is external (copied from a site where there is support and maintenance) or there is code internal to liboqs (i.e., maintained by the OQS team). "local upstream" thus sounds like something with no support/maintenance responsibility: Can you please clarify @praveksharma ?

[praveksharma]

What do you understand by "local upstream"? Either something is external (copied from a site where there is support and maintenance) or there is code internal to liboqs

The thought was that this and similar upstreams could be stored locally in a dir, say scripts/copy_from_upstream/local, with maintainence responsibility clearly documented in the CODEOWNERS file. @baentsch do you think that is sufficient deliniation of maintenance responsibility?

The reason I'd suggested this instead of a Nvidia repo is because Nvidia has a single monolithic repository for all their open source projects (is that correct @stevenireeves?). I'd assumed that would mean we may have to pull a repo of some significant size to use copy_from_upstream.py in the expected way; I have since then learned it is possible to do a "sparse git checkout" which would allow use to copy only specified directories. I shall look more into this.