With file system the resulting graph would be too big and too complicated to be useful. You would need heavy filtering to make any sense of it. In best case it would be a decent for determining where most of your files are, but those are probably hidden system directories and .git metadata.
Most tree visualization research is focused on static rendering of whole tree. For file browsing and many other applications it's better to have just section of tree with interactive navigation around trees. This makes visualization an easier with more predictable layout, and the user is not overwhelmed with data. This approach is basically replicating model of our attention/focus and how we make sense of complex structures.
The design of file browsers indeed seems to be stuck in 80s. Online storage services are copying this design as well. Is anyone aware of better design, in form of paper or prototype?