Non-blocking metadata extraction

_(Part of #23)_

We could reduce the initial wait time before folder opening by turning metadata extraction into a non-blocking process, similar to how we handle preview generation. Currently, a resource missing a preview is represented by a progress circle until the preview is ready. We should aim for the same approach with metadata.

Although, handling metadata can be slightly more complex given its intricate connection with the UI. For instance, depending on the chosen sorting criteria, the process of resource ordering can require metadata. So, if metadata isn't computed yet, we may not be able to sort resources in some cases. We need to thoroughly consider how metadata is intertwined with the UI and how to manage its absence correctly.

Start by tracing the existing non-blocking preview-generation flow and how metadata extraction, resource ordering, and sorting are connected to the UI. Done means folder opening no longer waits for metadata, missing metadata is represented appropriately, and sorting remains correct when metadata is not yet available.

Written by the indexing model from the issue text.