Extend diagnostic-timing-stats to optionally show counts

[lindig]

The Host.get_diagnostic_timing_stats API call shows mean and SD of time spent. We would like to know how many samples were used to this but can't simmply add this because it could break existing clients. Add an optional flag to include them with a default that maintains the existing behaviour.

The n=xxx is new and depends on the counts=true flag.

[root@eu1-dt034 ~]# xe  diagnostic-timing-stats counts=true | head
host-uuid                                                                      : 9eb655ef-3249-46fb-beab-d7d12c0acbaf
                                                                host-name-label: eu1-dt034
                                                         Stats reporting thread: 0.000136 [sd = 0.000000, n=1]
                                    Sync UEFI certificates on host with XAPI db: 0.012959 [sd = 0.000000, n=1]
                                                                   Db_gc: PGPUs: 0.000006 [sd = 0.000001, n=18]
                                                                Db_gc: Consoles: 0.000020 [sd = 0.000003, n=18]
                                       Cleanup attached pool_updates when start: 0.000003 [sd = 0.000000, n=1]
                                 Cancel_tasks.update_all_allowed_operations: SR: 0.000824 [sd = 0.000000, n=1]
                                                                   Db_gc: VTPMs: 0.000001 [sd = 0.000001, n=18]
                                               Initialise monitor configuration: 0.001076 [sd = 0.000000, n=1]