fix bind mount bug when using proxy

[trz42]

When a proxy is used to provide access to CernVM-FS on a compute node, we add the proxy configuration to /etc/cvmfs/default.local. The current code checked if this directory was already listed in the $BIND_PATHS, and exited if so.

However, exiting is too strict. We only should exit if the source of the existing bind mount with destination /etc/cvmfs/default.local is different to the currently changed source (where we're adding the proxy settings). If the sources are the same (we verify via readlink -f), we don't need to add the bind mount to $BIND_PATHS again (all is fine).

This PR adds the function check_bind_paths_for_target that checks whether a given src:target bind mount is already in $BIND_PATHS. This function is then used to decide whether the given bind mount has to be added or not OR if an ambiguous situation has occurred (different sources for same target) and thus eessi_container.sh shall terminate immediately.

Code was tested in trz42/software-layer#93 via https://github.com/trz42/software-layer-scripts/tree/debug_bind_mount_issue

[casparvl]

Tested here casparvl/software-layer#6 (comment) (even though I didn't have an issue with the same target being specified twice - at least it doesn't break any existing behavior). Looks good to me.