fix: simplify A2T return_timestamps type (#239)

In the previous pull request, I adhered to the union type used by Whisper, but
this caused issues in the multipart handling. To avoid complicating the MultiPartWriter,
this commit converts the union type to a string. This change simplifies the multipart
handling and user confusion.

runner/app/routes/audio_to_text.py

@@ -96,12 +96,13 @@ async def audio_to_text(
     ] = "",
     pipeline: Pipeline = Depends(get_pipeline),
     return_timestamps: Annotated[
-        Union[str, bool],
+        str,
         Form(
             description=(
                 "Return timestamps for the transcribed text. Supported values: "
-                "'sentence', 'word', or a boolean. Default is True ('sentence'). "
-                "False means no timestamps. 'word' means word-based timestamps."
+                "'sentence', 'word', or a string boolean ('true' or 'false'). Default "
+                "is 'true' ('sentence'). 'false' means no timestamps. 'word' means "
+                "word-based timestamps."
             )
         ),
     ] = "true",

runner/gateway.openapi.yaml

@@ -437,13 +437,12 @@ components:
           description: Hugging Face model ID used for transcription.
           default: ''
         return_timestamps:
-          anyOf:
-          - type: string
-          - type: boolean
+          type: string
           title: Return Timestamps
           description: 'Return timestamps for the transcribed text. Supported values:
-            ''sentence'', ''word'', or a boolean. Default is True (''sentence'').
-            False means no timestamps. ''word'' means word-based timestamps.'
+            ''sentence'', ''word'', or a string boolean (''true'' or ''false''). Default
+            is ''true'' (''sentence''). ''false'' means no timestamps. ''word'' means
+            word-based timestamps.'
           default: 'true'
       type: object
       required:

runner/openapi.yaml

@@ -448,13 +448,12 @@ components:
           description: Hugging Face model ID used for transcription.
           default: ''
         return_timestamps:
-          anyOf:
-          - type: string
-          - type: boolean
+          type: string
           title: Return Timestamps
           description: 'Return timestamps for the transcribed text. Supported values:
-            ''sentence'', ''word'', or a boolean. Default is True (''sentence'').
-            False means no timestamps. ''word'' means word-based timestamps.'
+            ''sentence'', ''word'', or a string boolean (''true'' or ''false''). Default
+            is ''true'' (''sentence''). ''false'' means no timestamps. ''word'' means
+            word-based timestamps.'
           default: 'true'
       type: object
       required: