feat: prevent premature namespace teardown and redundant re-deployments on worker restarts

[subhashkhileri]

https://issues.redhat.com/browse/RHIDP-12518

What

Moves namespace cleanup from the worker fixture's finally block to a custom Playwright reporter, and adds built-in protection against redundant deploy() calls when workers restart after test
failures.

Why

When a test fails, Playwright kills the worker and creates a new one for remaining tests. This caused:

• The fixture's finally to delete the namespace too early — breaking remaining tests
• beforeAll to re-run deploy() from scratch — wasting several minutes per failure

How

• deploy() now skips automatically if it already succeeded in the current test run (uses file-based flags via proper-lockfile)
• test.runOnce(key, fn) added for wrapping expensive pre-deploy setup (external services, scripts) that also shouldn't repeat
• TeardownReporter handles namespace deletion in onEnd() — runs in the main process, survives worker restarts, deletes after all tests complete
• registerTeardownNamespace() exported from ./teardown for consumers using custom namespaces

Consumer impact

No changes needed for the default configure() + deploy() pattern — works out of the box. For pre-deploy setup, wrap in test.runOnce():

// Simple — no changes needed
test.beforeAll(async ({ rhdh }) => {
await rhdh.configure({ auth: "keycloak" });
await rhdh.deploy();
});

// Pre-deploy setup — wrap in test.runOnce
test.beforeAll(async ({ rhdh }) => {
await test.runOnce("tech-radar-setup", async () => {
await rhdh.configure({ auth: "keycloak" });
await $bash deploy-service.sh ${rhdh.deploymentConfig.namespace};
await rhdh.deploy();
});
});