Because mount entries provide more information compared to just one
string mountpoint. For example later for `wcfs status` and `wcfs stop`
we will need to use the ID of a "device" that is attached to the mount,
and also the type of the filesystem that is serving the mount.

-> Introduce internal.os.MountDB to retrieve information from OS
   registry of mounted filesystems and use its entries instead of plain
   mountpoint string.

wcfs_test.py already had some rudimentary code to parse /proc/mounts
which we also replace with querying MountDB.

The API of MountDB might be viewed as a bit of overkill but it will
align with API of upcoming ProcDB for which it will be reasonable.