[Feature Request] Infer the test environment from the command varlock run launches #1145

Description

When varlock run launches a test runner, infer the test environment from the command itself, so VARLOCK_ENV is test without passing a flag.

Motivation

The Bun guide now recommends varlock run -- bun test. But varlock run resolves before the runner starts, and it's the runner that sets NODE_ENV=test, so VARLOCK_ENV comes out as development. Repro on 1.21.0:

# @defaultSensitive=false
# ---
ENV_NAME=$VARLOCK_ENV

// x.test.ts
import { test } from "bun:test";
test("env", () => console.log("ENV_NAME =", process.env.ENV_NAME, "NODE_ENV =", process.env.NODE_ENV));

varlock run -- bun test prints ENV_NAME = development NODE_ENV = test.

The builtin variables docs cover this and recommend passing the flag explicitly (VARLOCK_ENV=test varlock run -- vitest), so every test script in every package of a monorepo carries it. Forgetting it on one package runs its tests against development config without any error.

Proposed Solution

In varlock run, when VARLOCK_ENV is not already set, look at the command being run: bun test, vitest, jest, node --test, mocha (and the same through bunx/npx) resolve with VARLOCK_ENV=test. An explicit VARLOCK_ENV or @currentEnv value still wins. It could be an opt-in root decorator if guessing from argv is too implicit.

Alternatives

• Pass the flag in every test script (what the docs recommend today).
• A varlock test -- <cmd> shorthand that sets the test environment.

Additional Information

The same timing problem applies to anything that sets its own NODE_ENV after startup, but test runners are by far the common case.