Replace in-memory LC3 file loading with streaming byte generator to reduce memory usage

Refactors LC3 file handling to stream frames byte-by-byte from disk instead of loading the entire file into memory. Adds _lc3_file_byte_gen generator function that skips the 18-byte LC3 header, reads frame size headers, and yields individual bytes with optional looping. Removes read_lc3_file usage and itertools.cycle approach in favor of the new streaming generator.

src/auracast/multicast.py

@@ -551,6 +551,29 @@ async def init_broadcast(
     return bigs
 
 
+def _lc3_file_byte_gen(filename: str, loop: bool = False):
+    """Stream LC3 frames from disk as individual bytes, with optional looping.
+
+    Yields one byte (int) at a time so it is compatible with the existing
+    ``bytes(itertools.islice(gen, bytes_per_frame))`` consumer without loading
+    the whole file into memory.
+    """
+    while True:
+        with open(filename, 'rb') as f:
+            f.read(18)  # skip 18-byte LC3 header
+            while True:
+                size_b = f.read(2)
+                if len(size_b) < 2:
+                    break
+                frame_size = struct.unpack('=H', size_b)[0]
+                frame = f.read(frame_size)
+                if len(frame) < frame_size:
+                    break
+                yield from frame
+        if not loop:
+            return
+
+
 class Streamer():
     """
     Streamer class that supports multiple input formats. See bumble for streaming from wav or device
@@ -706,13 +729,7 @@ class Streamer():
                 big['precoded'] = True
                 big['lc3_bytes_per_frame'] = global_config.octets_per_frame
                 filename = big_config[i].audio_source.replace('file:', '')
-
-                lc3_bytes = read_lc3_file(filename)
-                lc3_frames = iter(lc3_bytes)
-
-                if big_config[i].loop:
-                    lc3_frames = itertools.cycle(lc3_frames)
-                big['lc3_frames'] = lc3_frames
+                big['lc3_frames'] = _lc3_file_byte_gen(filename, loop=big_config[i].loop)
 
             # use wav files and code them entirely before streaming
             elif big_config[i].precode_wav and big_config[i].audio_source.endswith('.wav'):