Fix clients mount for non-mounted compose clusters

[csegarragonz]

Currently, even if we don't mount the source code for Faasm, we still mount the clients code (i.e. clients/cpp and clients/python) from ./clients/cpp and ./clients/python). I link the offending lines.

This has a domino effect:

1. Even if not mounting the source code, we need to clone faasm with all the submodules. Theoretically, we could get away with just wget-ing the docker-compose.yml file, but that's not necessary neither. I think one shallow clone from the tag without submodule is enough (and can also be shared with k8s deployments).
2. Even if not mounting the source code, we need to wait for the venvs in client/{cpp,python}/venv/faasm_venv.BUILT.
3. Even if we bump the faasmctl version in clients/cpp (and bump the container tag to have faasmctl inside the contianer), we also need to bump the main faasm tag.

[csegarragonz]

The solution would be to remove the mount points from the docker-compose.yml file, and pass them as parameters from the command line in faasmctl (tasks/deploy.py).