Optimize cirque bootstrap time

[yunhanw-google]

Problem

Currently cirque bootstrap(build and compile dockers) is time-consuming, 20 minutes, we should upload chip-cirque-device-base image to dockerhub, and download it when running test.

Also currently we create 6 docker node based upon chip-cirque-device-base for on-off test, echo-test, on-off cluster test,
we should be able to just bring up two docker node, run tests on them.

TODO:

• 1. Upload basic chip-cirque-device-base image to docker hub
• 2. Build openthread only when needed