dev/Makefile MetalLB config fails on dual-stack (IPv6+IPv4) Docker networks

Problem

The dev/Makefile uses this approach to extract the MetalLB IP range:

```makefile
KIND_NET_CIDR=$(docker network inspect kind -f '{{range .IPAM.Config}}{{.Subnet}}{{"\n"}}{{end}}' | grep '.' | head -n1)
```

This works on systems where the kind Docker network has only IPv4, or where IPv4 appears first. However, on systems with dual-stack Docker networks (e.g., Fedora 43 with Docker's default IPv6 support), the network may return IPv6 subnets first:

```
fc00:f853:ccd:e793::/64
172.19.0.0/16
```

The current template Makefile already handles this correctly with the `grep '\.'` filter. This is just a note that this is a real issue we hit during testing - the filtering is important and should not be removed.

Context

When building the nebari-superset-pack, the initial Makefile didn't include the IPv4 filter, causing MetalLB configuration to fail with an invalid CIDR error.

Start with dev/Makefile and compare its KIND_NET_CIDR extraction with the current template Makefile. Verify how Docker dual-stack output is handled and confirm that the IPv4 subnet is selected for MetalLB; done means the Makefile remains valid when IPv6 appears first.