e2e: auto-install pinned Node from .nvmrc when the local version drifts

While setting up the dedicated e2e-automation machine, the box had Node `v24.14.0` but the repo pins `v24.18.0` in `.nvmrc`. Setup required a manual `nvm install`.

The trusted-lane runner (`e2e/automation/run-trusted.sh`) sources nvm and runs `nvm use`, but swallows failure (`|| true`) and does **not** install a missing version — so on version drift it silently runs the wrong Node (or the pinned version is simply absent) instead of self-healing.

**Proposal:** before `nvm use`, run `nvm install` (reads `.nvmrc`, installs+uses the pinned version, no-op if already present) so the autonomous machine self-heals when Node drifts rather than failing or running an unpinned Node. Applies to the harness run scripts that drive the suite.

_Context: surfaced during dedicated-machine setup for the Studio e2e automation._

Start by reading e2e/automation/run-trusted.sh and identify the other harness run scripts that drive the suite. Check how they source nvm, read .nvmrc, and invoke nvm use; done means the pinned Node version is installed and selected after version drift, so the trusted-lane run needs no manual setup.

Written by the indexing model from the issue text.