Python runtime: completed processes not getting cleaned up

[varunbpatil]

In the case when the function completes successfully or completes with an exception (i.e, non-timeout path), the processes are not cleaned up due to a missing join leading to processes getting accumulated over the course of the pod lifetime and resulting in the node running out of PIDs to assign to new processes.

I have raised a PR #82 for the same.

It might even be a good idea to use the tini init as the ENTRYPOINT for the Docker images so that even zombie processes created by misbehaving user code can be reaped. I'd be happy to submit a PR for that.

cc: @andresmgot