Add lab test harness for bulk brick lifecycle and NX-OS BGP cleanup

[elinalin]

Summary

• Add `delete` command to the lab test harness (`test/lab/main_test.go`) so test cases can delete Kubernetes resources and wait for finalizers to complete
• Add `SSH_PORT`, `GNMI_PORT`, `GNMI_ADDR`, and `NAMESPACE` environment variable support to allow port-forwarding and namespace isolation when running against clabernetes
• Add `ResolveAddr` helper to resolve hostnames to IPv4 (required by the Device address field)
• Set NX-OS provider name on the Device created by the test harness
• Delete pre-existing objects in `Create()` to recover cleanly from interrupted test runs
• Increase reconciliation timeout from 10s to 60s for real-device testing
• Add precondition check to BGP test cases: verify no `router bgp` config exists before the test begins
• Add test case 4: NX-OS default BGP instance create and delete (`bgp_default.txt`, `bgp.txt`)
• Add test case 5: NX-OS per-VRF BGP selective delete (`bgp_vrf.txt`, `bgp.txt`)
• Add lab runbook for bulk brick lifecycle validation (`docs/developer-guide/bulk-brick-lifecycle-lab.md`)

Test results

Test case 4 (default BGP create + delete): PASS — global `router bgp` instance and `operator-managed--default` marker fully removed after deletion.

Test case 5 (per-VRF BGP selective delete): documents a known bug in `deleteBGP` introduced in `1c03adb8`. After deleting both VRF-scoped BGP bricks, `router bgp 65000` and ownership markers remain on the device. Root cause: a post-deletion `GetConfig` reads stale state due to NX-OS's gNMI consistency window and incorrectly concludes a managed domain still exists. This test case is intentionally included to reproduce the bug — it is expected to fail until the bug is fixed separately.

Test plan

• Run `bgp_default.txt` (test case 4) against a real NX-OS device and confirm PASS
• Run `bgp_vrf.txt` (test case 5) against a real NX-OS device — step 3 is expected to FAIL (known bug, tracked separately)