A direct implementation using DriveInfo directly on a path does not work
as expected. The method will return a DriveInfo object with the given
path as both the Name and the RootDirectory, which is not helpful.
Instead, add parsing logic to find the best possible match out of all
filesystems on the system for the path, including handling edge cases
involving symlinked paths in the chain.
This ensures that the resulting DeviceId is a valid filesystem, allowing
it to be used in the UI to show a better description. It also includes
the new ResolvedPath which will show, if required, what the Path
resolved to after all symlinks are interpolated.
One possible issue here is that walking all drives as-is might become
slow(er) on a system with many partitions, but even on my
partition-heavy system with over a dozen ZVOLs and remote mounts, this
takes under 0.4 seconds including runup time for `dotnet run`, so I
imagine this should be fine.
Don't enforce a minimum amount of free space for the tmp and log dirs
Original-merge: 097cb87f6f
Merged-by: crobibero <cody@robibe.ro>
Backported-by: Bond_009 <bond.009@outlook.com>
