Per-OSD dashboard should include read and write latency

Something like the following:

`irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000
`

`irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000
`

It might also be nice to have a global latency view on the homepage (perhaps merged w/ the apply vs commit latency graph?).


Start by locating the dashboard definition and its existing per-OSD panels, then compare the available Ceph read and write latency metrics with the two PromQL expressions in the issue. Done means the per-OSD dashboard shows read and write latency; determine separately whether the optional global homepage view is in scope.

Written by the indexing model from the issue text.