Chunked decryption, resume & selection fixes

Implement chunked decryption of .m4s segments (avoid loading entire stream into memory) by changing decrypt() to accept a segment directory, process ~1GB chunks with mp4ff-decrypt, and concatenate results. Add resume support in segment downloader by skipping already downloaded .m4s files and initializing successful counter accordingly. Improve manifest handling: gracefully handle missing AVC/HEVC tracks, bruteforce ticks conditionally, and mark representations for removal when no valid tick is found. Add fallback unfiltered selection in select_track when filtered lookup yields no candidates. Replace naive UTC offset logic with explicit Europe/Paris timezone utilities and adjust related time conversion signatures. Bump project version to 1.0.

utils/stream.py

@@ -456,8 +456,19 @@ async def save_segments(
 
     # Build list of all segments to download
     segments_to_download = [(start_tick + i * duration, i) for i in range(rep_nb)]
+
+    # In case of resuming, check which segments are already downloaded
+    already_downloaded = [
+        int(f.split(".")[0])
+        for f in os.listdir(f"{output_folder}/segments_{track_id}")
+        if f.endswith(".m4s") and f.split(".")[0].isdigit()
+    ]
+    segments_to_download = [
+        (tick, rep) for tick, rep in segments_to_download if tick not in already_downloaded
+    ]
+
     retry_list = []
-    successful = 0
+    successful = len(already_downloaded)
 
     async with aiohttp.ClientSession() as session:
         # Process segments in batches