Transcription: dwindle model choice to Whisper Base + Whisper Large v3

Registry now lists exactly two models instead of five:
- 'Whisper Base' — names the actual bundled model (was 'Base (default)').
- 'Whisper Large v3' — kept; honest copy on size/speed/RAM.
tiny/small/medium removed from the registry (validate_model_name now
rejects them; existing rows keep their saved model — all current data
is 'base', still valid).
Detail screen shows the model that WILL run by real display name
('Whisper Base (default)') instead of 'Use default (base)'; the picker
lists the two models with the default marked inline.
Backend 78 passed/1 skipped; app 34/34; engine hot-reloaded clean.

backend/shonar/services/ai/model_registry.py

@@ -37,42 +37,20 @@ class TranscriptionModelInfo:
 
 
 SUPPORTED_TRANSCRIPTION_MODELS: dict[str, TranscriptionModelInfo] = {
-    "tiny": TranscriptionModelInfo(
-        name="tiny",
-        display_name="Tiny",
-        description="Fastest and lightest. Good for quick drafts and slow machines.",
-        params="~39M",
-        approx_memory="~1 GB RAM",
-        relative_speed="~10x real-time (CPU)",
-    ),
     "base": TranscriptionModelInfo(
         name="base",
-        display_name="Base (default)",
-        description="Balanced default. Works reasonably well on ordinary computers.",
+        display_name="Whisper Base",
+        description="The bundled model — works offline out of the box. "
+        "Good for clear speech on ordinary computers.",
         params="~74M",
         approx_memory="~1 GB RAM",
         relative_speed="~7x real-time (CPU)",
     ),
-    "small": TranscriptionModelInfo(
-        name="small",
-        display_name="Small",
-        description="Better accuracy with higher resource usage.",
-        params="~244M",
-        approx_memory="~2 GB RAM",
-        relative_speed="~4x real-time (CPU)",
-    ),
-    "medium": TranscriptionModelInfo(
-        name="medium",
-        display_name="Medium",
-        description="Higher accuracy and slower performance.",
-        params="~769M",
-        approx_memory="~5 GB RAM",
-        relative_speed="~2x real-time (CPU)",
-    ),
     "large-v3": TranscriptionModelInfo(
         name="large-v3",
-        display_name="Large v3",
-        description="Highest accuracy and greatest resource requirements.",
+        display_name="Whisper Large v3",
+        description="Best accuracy (accents, names, messy audio). Downloads "
+        "~3 GB once; slow on CPU — best on machines with 16 GB+ RAM.",
         params="~1.5B",
         approx_memory="~10 GB RAM",
         relative_speed="~1x real-time (CPU)",