rmlock: Micro-optimize read locking

[markj]

Summary

Use get_pcpu() instead of an open-coded pcpu_find(td->td_oncpu). This
eliminates some unneccessary memory accesses and on amd64 results in a
shorter instruction sequence. Note that get_pcpu() didn't exist when
rmlocks were added.

[mjg]

please mfc this to stable/12