balance scheduler service number when start scheduler

[sonyafenge]

Part of issue: #120. after this fix, scheduler has balanced number of service on each machine, but the RegisterClientDuration still have big difference:

file name	RegisterClientDuration
grs-down-dismt-client-us-central1-a-mig-f361.log.0	365.287036ms
grs-down-dismt-client-us-central1-a-mig-f361.log.1	521.043594ms
grs-down-dismt-client-us-central1-a-mig-f361.log.2	342.378137ms
grs-down-dismt-client-us-central1-a-mig-f361.log.3	368.195141ms
grs-down-dismt-client-us-central1-a-mig-f361.log.4	360.4061ms
grs-down-dismt-client-us-central1-a-mig-f361.log.5	315.15387ms
grs-down-dismt-client-us-central1-a-mig-f361.log.6	320.079994ms
grs-down-dismt-client-us-central1-a-mig-f361.log.7	320.47048ms
grs-down-dismt-client-us-central1-a-mig-f361.log.8	374.838644ms
grs-down-dismt-client-us-central1-a-mig-f361.log.9	313.702576ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.0	5.46519ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.1	6.91962ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.2	6.453254ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.3	5.610316ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.4	5.197985ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.5	5.04598ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.6	5.100392ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.7	5.334791ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.8	5.047824ms
grs-down-dismt-client-us-central1-a-mig-z8mf.log.9	5.13324ms

[yb01]

this pr is just for fixed number ( or ensure they are evenly assigned on client machines). we will keep the issue open for investigating the registration duration differences. seems one machine is much slower than they other one

[yb01]

/lgtm