Pytest runs out of memory when coverage is enabled

[mint570]

This is found in #2237. The last few pytest failed.
We suspect that the cause is the coverage enabling: https://github.com/Azure/sonic-swss/blob/bf4d89082ad413ef71eb40a8ff2f279b87f1b002/.azure-pipelines/test-docker-sonic-vs-template.yml#L72

The "--keeptb" flag in pytest will keep the docker containers alive after the test finishes. In local experiments, each test will open two containers. After all the tests finish, there will be hundreds of containers open. If we add more pytest tests, the workflow will run out of memory.

We should try to run one test at a time. Collect the coverage artifacts, kill the containers. And then begin the next test.

[mint570]

@pettershao-ragilenetworks
@donNewtonAlpha

[mint570]

@bocon13 had mentioned some good approaches to consider:

1. Mount the file system instead of keeping the container running. So that we can collect the code coverage files even after the container finishes.
2. Keep the container file system but stop the compute part of the vs docker to reduce resource usage when pytest is running.

But in both the approaches, we will still need to do something at the end of each test (send a kill signal, collect/rename the files etc). Maybe that can be done as part of the pytest workflow?