test/nvme: Pick max 2 nvmes for sw_hotplug tests

Under our CI we have phy nodes with multiple nvmes, some with 3 and
more. This is how debug (introduced via 9dad1237) looks like under
the sw_hotplug suite on the node with 3 nvmes (per test):

[2023-12-12T18:51:34.632Z] remove_attach_helper took 102.18s to complete (handling 3 nvme drive(s))
[2023-12-12T18:53:54.349Z] remove_attach_helper took 121.37s to complete (handling 3 nvme drive(s))
[2023-12-12T18:55:54.476Z] remove_attach_helper took 120.45s to complete (handling 3 nvme drive(s))

This is how it looks under node with single nvme:

p2023-12-12T21:41:46.306Z] remove_attach_helper took 31.52s to complete (handling 1 nvme drive(s))
[2023-12-12T21:42:46.143Z] remove_attach_helper took 49.47s to complete (handling 1 nvme drive(s))
[2023-12-12T21:43:35.301Z] remove_attach_helper took 49.29s to complete (handling 1 nvme drive(s))

The difference is quite significant - the first case requires over 6m
to finish while second ~2.3m (this accounts the total time entire
sw_hotplug suite needed to complete).

The former case started triggering timeouts under nvme-phy-autotest
job at the very end of it - so those couple of minutes actually
matter.

To mitigate, limit number of nvme drives picked for this test to
max to save some time.

Change-Id: Ia55715df055662344a04f5cf45245c80d5086745
Signed-off-by: Michal Berger <michal.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/21029


Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <jim.harris@samsung.com>