[Tests][Serve] Smoke test on SkyServe has a flacky output check

[cblmemo]

In our SkyServe related tests, there are a lot of commands to check the number of replicas:

+ s=$(sky serve status t-ss-new-autosca-6c-402b-f1-blue_green); until ! echo "$s" | grep "Controller is initializing."; do echo "Waiting for serve status to be ready..."; sleep 5; s=$(sky serve status t-ss-new-autosca-6c-402b-f1-blue_green); done; echo "$s"; echo "$s";  echo "$s" | grep "(\[Spot\]vCPU=2)" | grep "PROVISIONING\|STARTING\|READY" | wc -l | grep 4 || exit 1; echo "$s" | grep "(vCPU=2)" | grep "PROVISIONING\|STARTING" | wc -l | grep 1 || exit 1; echo "$s" | grep "(vCPU=2)" | grep "READY" | wc -l | grep 2 || exit 1;

where it greps the vCPU=2 and count the number of lines. However, it is possible that the cluster is not launched yet, and the resoruces is still shown as -:

Services
NAME                                    VERSION  UPTIME  STATUS  REPLICAS  ENDPOINT            
t-ss-new-autosca-6c-402b-f1-blue_green  1        2m 47s  READY   2/7       34.67.113.93:30001  

Service Replicas
SERVICE_NAME                            ID  VERSION  ENDPOINT                    LAUNCHED        RESOURCES             STATUS        REGION          
t-ss-new-autosca-6c-402b-f1-blue_green  1   1        http://54.204.214.237:8081  2 mins ago      1x AWS(vCPU=2)        READY         us-east-1       
t-ss-new-autosca-6c-402b-f1-blue_green  2   1        http://54.175.176.31:8081   2 mins ago      1x AWS(vCPU=2)        READY         us-east-1       
t-ss-new-autosca-6c-402b-f1-blue_green  3   2        -                           -               -                     PROVISIONING  -               
t-ss-new-autosca-6c-402b-f1-blue_green  4   2        -                           a few secs ago  1x AWS([Spot]vCPU=2)  PROVISIONING  ap-northeast-2  
t-ss-new-autosca-6c-402b-f1-blue_green  5   2        -                           -               -                     PROVISIONING  -               
t-ss-new-autosca-6c-402b-f1-blue_green  6   2        -                           -               -                     PROVISIONING  -               
t-ss-new-autosca-6c-402b-f1-blue_green  7   2        http://54.159.194.181:8081  28 secs ago     1x AWS(vCPU=2)        STARTING      us-east-1

And the tests immediately failed and exited. This is likely to happen when large scale tests is running in parallel, and the cluster might stay at PENDING state for a while, delaying the provision.

Should we wait for a while for it to turn into the resoruces representation?