Fallback to stream-based device discovery when no input/output port is available

### Problem

My Focusrite Clarett 4Pre USB is not listed in elementary Sound’s **Input** tab, even though it is active and usable.

After debugging between `pactl`, testing on elementary, pavucontrol and GNOME, here is the behavior on my machine:

- Clarett appears in **Output** in elementary Sound
- Clarett does **not** appear in **Input** in elementary Sound
- Clarett input/output are both visible and selectable in GNOME Sound and pavucontrol
- Default source is already the Clarett pro input (alsa_input...pro-input-0)
- Device is on pro-audio profile

I have checked the code, and this seems related to how devices are discovered:

- elementary currently relies heavily on **card ports** and active-port matching to build/bind device rows
- in this Clarett pro-audio case, the input stream exists but does not expose usable input port metadata (`Ports` / `Active Port`)
- result: no input device row is created for Clarett in elementary

### Proposal

Keep current port-based discovery, but add a fallback path that discovers devices from real source/sink streams when no matching port-backed device exists.

Suggested behavior:

1. Continue using card ports as primary discovery for normal hardware.
2. If a valid non-monitor source/sink exists but no port-backed row can represent it, create a stream-backed device row.
3. Apply this for both Input and Output.
4. Deduplicate when a port-backed and stream-backed row refer to the same endpoint.

### Prior Art (Optional)

I have no idea how hard the implementation could be, but at least I see that both GNOME and pavucontrol handle it.

GNOME Sound appears to handle this via `gvc`.

And `pavucontrol` enumerates sinks/sources directly from PulseAudio, and ports are only attached as optional metadata. If no ports, it just hides port controls, but the device row itself remains visible.