[DGX Station][Install] Re-running v0.0.97 Express install rebuilds the CLI twice and takes 154 seconds

## Description

Re-running the v0.0.97 Express installer on an already healthy dual DGX Station deployment preserves the running vLLM service, but repeats the CLI source clone, dependency installation, build, and link path twice. It also creates a sandbox backup and takes 154 seconds.

Platform scope: Reproduced on dual DGX Station only; other platforms not tested

Regression: Unknown — earlier versions not tested

## Environment

```text
Device: Dual DGX Station
OS: Ubuntu 24.04.4 LTS
Architecture: aarch64
Node.js: v22.23.2
npm: 10.9.8
Docker: 29.2.1
OpenShell CLI: 0.0.85
NemoClaw: v0.0.97
OpenClaw: 2026.7.1 (2d2ddc4)
```

## Steps to Reproduce

1. Prepare two DGX Stations using the supported dual-Station fabric setup.
2. From Station A, run:

```bash
curl -fsSL https://www.nvidia.com/nemoclaw.sh | \
NEMOCLAW_INSTALL_TAG=v0.0.97 \
NEMOCLAW_DGX_STATION_PEER={username}@{station-b-hostname} \
bash
```

3. Confirm the distributed vLLM service is healthy and a NemoClaw sandbox is registered.
4. Run the exact same Express install command again without changing the requested tag or peer.
5. Measure the elapsed time and observe the installer phases.

## Expected Result

When the requested tag is already installed and healthy, the installer should reuse the existing CLI artifact and perform only the validation or reconciliation work needed for that state. It should not clone, build, and link the same CLI twice.

## Actual Result

The running vLLM service remains available, but the same-version path clones, installs dependencies, builds, and links the CLI twice, creates a sandbox backup, and takes 154 seconds.

## Logs

```text
Not captured
```