MINOR: task: align the rq and wq locks

We really don't want them to share the same cache line as they are
expected to be used in parallel. Adding a 64-byte alignment here shows
a performance increase of about 4.5% on task-intensive workloads with
2 to 4 threads.
1 file changed