fix: add frame buffering to NoAudioProcessor for ESP32-C5 (#1766)

* add: Moji 2 has built-in ESP32-C5 dual band Wi-Fi

* feat: PowerSaveTimer 160 >> 240

* fix(audio): add frame buffering to NoAudioProcessor

On chips where CONFIG_USE_AUDIO_PROCESSOR is unavailable (e.g., ESP32-C5,
which is excluded by Kconfig despite having PSRAM), NoAudioProcessor is used.
The original implementation passed through 160-sample chunks directly, but
the Opus encoder expects 960 samples (60ms @ 16kHz), resulting in frame size
mismatch errors, high CPU usage, and watchdog resets.

Added output_buffer_ to accumulate audio samples until reaching
frame_samples_ (960) before outputting, matching AfeAudioProcessor behavior.

other:
Turn off sleep mode (PowerSaveTimer)

main/audio/processors/no_audio_processor.h

@@ -26,6 +26,7 @@ public:
 private:
     AudioCodec* codec_ = nullptr;
     int frame_samples_ = 0;
+    std::vector<int16_t> output_buffer_;
     std::function<void(std::vector<int16_t>&& data)> output_callback_;
     std::function<void(bool speaking)> vad_state_change_callback_;
     std::atomic<bool> is_running_ = false;