Lower num_pcie_ports to 12 in the nova-next job

This change is being proposed in response to guest kernel panics that
are happening occasionally in the nova-next CI job. In these particular
cases, we see messages like the following in the guest console log dump
[1]:

  pci 0000:00:03.6: BAR 13: no space for [io  size 0x1000]
  pci 0000:00:03.6: BAR 13: failed to assign [io  size 0x1000]

which seems to indicate the guest is running out of memory to allocate
to the pcie devices.

This lowers num_pcie_ports to 12 from 24 in an effort to use less
memory.

[1] https://pastebin.com/raw/cVDem0Xr

Change-Id: Ibfa65a00dd91e6c3f28fef837a7e172221f72f42

Author: melwitt

.zuul.yaml

@@ -411,7 +411,7 @@
             libvirt:
               # Increase the number of PCIe ports per instance given the q35
               # machine type attaches more devices by default than pc
-              num_pcie_ports: 24
+              num_pcie_ports: 12
               hw_machine_type: "x86_64=q35"
               # Use lower TB cache than default(1GiB), only applicable with
               # libvirt>=8.0.0