chore(gpu): rework ci to adapt to the shortage of h100

[agnesLeroy]

closes: please link all relevant issues

PR content/description

I'm removing classic PBS tests from the H100 workflows and moving them to dedicated workflows on RTX to be sure we test them, since there's currently a shortage of H100's on Hyperstack.

Check-list:

• Tests for the changes have been added (for bug fixes / features)
• Docs have been added / updated (for bug fixes / features)
• Relevant issues are marked as resolved/closed, related issues are linked in the description
• Check for breaking changes (including serialization changes) and add them to commit message following the conventional commit specification

[soonum]

Good workaround 👍
I'm curious, why do we experience H100 shortage ? This chip is already too old for Hyperstack ?

[agnesLeroy]

@soonum no I think it's just the platform has more & more clients. They will increase their capacity but I have no timeline about this atm. Blackwells won't be available before at least a year for the on-demand offer, because Nvidia had delays on it.