ci[gpu]: run cuda micro-benchmarks with codspeed

[0ax1]

No description provided.

[0ax1]

Giving this a couple of runs first to check whether the benches are stable across runs.

[codspeed-hq]

Merging this PR will not alter performance

✅ 1130 untouched benchmarks
⏩ 33 skipped benchmarks¹

---

_{Comparing ad/gpu-codspeed (8d16718) with develop (140eec6)}

Footnotes

1. 33 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩