fix: use larger disks to avoid throttling

[lenaschoenburg]

As we saw in camunda/camunda#21758, a 32 GiB disk is slightly too small for our default workload.

[lenaschoenburg]

@Zelldon Ping on this one. Let me know if you disagree.

[ChrisKujawa]

Currently unsure whether we should increase the disk size as it increases cost (not much i guess but still).

I'm currently investigating whether we can move our benchmarks to a different process model/load, which might not suffer of this.

[npepinpe]

We have more budget now since we dropped the whole CI runners 👍

I'd rather we reduce work by throwing money at it, and correct afterwards (we do keep track of usage).