Improper test cleanup and order-dependent results in long tests

[kaitozaw]

The long test suite (rake long_only) has two reliability issues:

1. Orphaned BeEF Processes After Test Runs
   When tests finish (or fail), the forked BeEF server processes are not reliably killed, causing subsequent test runs to fail with:Port 3000 is already in use. Exiting.

2. Seed-Dependent Test Failures
   Test results are flaky and depend on the execution order (controlled by RSpec seed).