perf test env setup enhancement works

[yb01]

This issue intends to capture needed improvements in perf test tools, namely, kube-up, kubemark, clusterloader etc. for Arktos 430 or beyound.

1. reduce test env setup time: such as parepare VMs in parallel, start hollow-nodes in parallel, cleanup VMs in parallel
2. variable naming consistency
3. make clear cut for AWS cloud providers -- currently some logic are in start-kubemark.sh which can be moved to AWS related libs

[sonyafenge]

these two files need be deleted when run stop-kubemark.sh, this will caused start-kubemark failed if already existing.
/tmp/haproxy_cfg_generator
/tmp/saved_tenant_ips.txt

[sonyafenge]

Detailed design information for this enhancement attached:
kubeup-kubemark.script.pptx

[sonyafenge]

Steps and plans for this enhancement tasks:

• Move proxy setup logic from dev machine to proxy; Move proxy setup logic from dev machines to proxy VM #1212
• There are part of setup logic for TP/RP in dev machine, move all these logic from dev machine to TP/RP server; kube-up.sh support scaleout #1227
• Both of kube-up and kubemark support scale-out structure; kube-up.sh support scaleout #1227
• Cleanup all cloud provider independent logic and move these logic to cluster/common.sh
• Cleanup proxy setup and make sure it is independent with kubernetes/arktos setup

[sonyafenge]

Keep this issue opened, still working on related items.