Prevents loaded packages miscount (e.g. -1) in dashboard/log in sandbox
instance.
This occurs because the benchmark line uses a simple heuristic to
determine the loaded packages: length of load-path minus the length of
doom--initial-load-path (to save on more expensive counting methods).
However, in the sandbox, load-path is pre-populated with all packages
right from the get-go, so doom--initial-load-path will be incorrect.