D_v7 VMs have improper socket/core/thread spec

CycleCloud Version: 8.9.0-3754
Cluster: ccw (version 8.9.x)

**Observed:**
`sinfo -o "%.9P %.5D %.4c %.7X %.5Y %.7Z"` shows 1 thread per core for a node array configured with Standard_D16as_v7 :
```
PARTITION NODES CPUS SOCKETS CORES THREADS
hpc* 6 16 16 1 1
```
While `srun lscpu`: `Thread(s) per core: 2`.

**Expected:**
**EITHER** threads per core are properly shown as 2 and cons_tres handles 8 cores per node, same as e.g. for D16as_v5:
```
PARTITION NODES CPUS SOCKETS CORES THREADS
hpc* 5 16 8 1 2
```
**OR** disable SMT so that for D16as_v7 sinfo shows 8 CPU per node, 1 thread per CPU.

No source file or test is named. Start by reproducing the Standard_D16as_v7 output with sinfo, srun lscpu, and the cons_tres configuration, then compare it with the D16as_v5 example. Done means the reported socket/core/thread values and CPU allocation consistently match one of the two expected configurations.

Written by the indexing model from the issue text.