feat: add audio transcription pipeline for voice messages

Adds Whisper-compatible audio transcription via configurable endpoint.
New functions: isSupportedAudio(), mimeToExtension(), transcribeAudio(),
buildUserMessageWithAudio(). Config schema gains audio section with
transcription_endpoint, api_key, and model. Daemon wires transcription
into the message router. Channel adapters extract audio from voice/audio
messages (Telegram voice+audio, Discord audio/*, Slack audio/*, WhatsApp
ptt+audio). Includes 57 media tests (was 25, now covers all audio paths).

src/config/schema.ts

@@ -180,6 +180,12 @@ const webSearchSchema = z.object({
   max_results: z.number().min(1).max(20).default(5),
 }).default({});
 
+const audioSchema = z.object({
+  transcription_endpoint: z.string().optional(),
+  transcription_api_key: z.string().optional(),
+  transcription_model: z.string().default('whisper-1'),
+}).default({});
+
 // ── Tool policy schemas ──────────────────────────────────────────────
 
 const toolProfileEnum = z.enum(['minimal', 'messaging', 'coding', 'full']);
@@ -259,6 +265,7 @@ export const configSchema = z.object({
   browser: browserSchema,
   retry: retrySchema,
   web_search: webSearchSchema,
+  audio: audioSchema,
   prompt: promptSchema,
   tools: toolsSchema,
   sandbox: sandboxSchema,
@@ -274,6 +281,7 @@ export type AgentsConfig = z.infer<typeof agentsSchema>;
 export type CompactionConfig = z.infer<typeof compactionSchema>;
 export type MemoryConfig = z.infer<typeof memorySchema>;
 export type WebSearchConfig = z.infer<typeof webSearchSchema>;
+export type AudioConfig = z.infer<typeof audioSchema>;
 export type ProcessConfig = z.infer<typeof processSchema>;
 export type BrowserConfig = z.infer<typeof browserSchema>;
 export type DiscordConfig = z.infer<typeof discordSchema>;