Parallel workers need a Redis database each

A parallel suite gets a database per worker automatically and shares every other backing service, which produces failures that look exactly like ordering bugs.

$token = ParallelTesting::token();

config([
    'cache.prefix' => "test_{$token}_",
    'database.redis.default.database' => (int) $token,
    'filesystems.disks.local.root' => storage_path("testing/{$token}"),
    'queue.connections.redis.queue' => "test_{$token}",
]);

Redis is the one that catches people because a shared cache across eight workers produces intermittent failures with no pattern. Finding all of the shared resources takes a couple of runs, and keeping the configuration in one place with a comment explaining why is what stops somebody adding a ninth without noticing.