From toil worker image, abstract out shared worker filesystem

[theferrit32]

As a prerequisite for using toil image, require a shared filesystem be set up on all worker nodes, that the toil-workspace used to launch toil can also access.

When toil worker image is run, either mount the filesystem from the worker node as a volume, or accept some sort of argument that tells the container itself what shared filesystem to mount.

Then when toil is run, use the path to that filesystem mount as the --jobStore value, instead of the /renci/irods mount.

[theferrit32]

Launcher and worker containers now work with a hardcoded NFS host mountpoint as an automatically mounted volume, using btrfs as the backing filesystem on the server node. Will look into making this configurable as a parameter. May investigate performance tweaks or other protocols.

[dcvan24]

Btrfs seems only working on a single node with multiple block devices, but cannot be shared across multiple nodes AFAIK. I think a distributed storage system such as Ceph, Lustre and ZFS may be a better fit for this use case. It needs investigations into these systems to make a wise selection for our environment.