add configuration profiles with per-invocation --profile flag

- Add SystemPrompt field to HushConfig (empty = built-in default)
- Refactor ConfigManager: extract ApplyTomlFields, add LoadWithProfile(),
  ListProfiles(), GetProfilePath(), EnsureProfilesDirExists(); remove
  HUSH_PROFILE env-var logic (profiles are now resolved by the CLI)
- Extend socket protocol: action commands (START/STOP/TOGGLE/ABORT) now
  carry a [4-byte LE length][optional HushConfig JSON] payload so the CLI
  can pass a per-invocation config override without restarting the daemon
- Add GENERATE_PROFILE (cmd 7) socket command: CLI sends a description,
  daemon calls the LLM and returns a generated system prompt
- Orchestrator: StopAndProcessAsync accepts optional HushConfig override;
  ProcessWithLlmAsync uses proper system/user chat roles and respects
  config.SystemPrompt; add GenerateProfilePromptAsync
- Split CompleteTextAsync signature to (systemPrompt, userMessage, model)
  across ITextStreamingProvider, GroqProvider, FireworksProvider
- Add --profile/-p flag to hush toggle and hush stop
- Add hush profiles subcommand: list, get, new (manual or AI-generated), edit

Hush.Providers/src/Interfaces/IAudioToTextProvider.cs

@@ -15,5 +15,6 @@ public interface IAudioToTextProvider
     Task<string> TranscribeAsync(
         Stream audioStream,
         string modelName,
+        string? language = null,
         CancellationToken cancellationToken = default);
 }