openhcl_servicing_keepalive test fails with sidecar enabled

[jstarks]

Probably this is due to some kind of memory corruption caused by sidecar memory
conflicting with keepalive memory, or something like that. We'll disable sidecar
for this test until this can be fixed, since keepalive is not production ready
anyway.

[chris-oo]

I suspect the issue is that the dma gpa range is being stomped on by the servicing load & nvme device. I'll see if i can take a look later this week.

[chris-oo]

This actually looks like some kind of race between an interrupt arriving and us onlining a sidecar VP. See #2070, but we fail on every test with this assertion:
openhcl/underhill_core/src/vp.rs:300:13: