web/workers/fetch: catch network-related errors & retry 3 times

previously all network issues showed a "worker crashed" error, which people misinterpreted all the time, and reasonably so

web/i18n/en/error/queue.json

@@ -7,6 +7,7 @@
     "fetch.no_file_reader": "couldn't write a file to cache",
     "fetch.empty_tunnel": "file tunnel is empty, try again in a few minutes",
     "fetch.corrupted_file": "file wasn't downloaded fully, try again",
+    "fetch.network_error": "downloading was interrupted by a network issue",
 
     "ffmpeg.probe_failed": "couldn't probe this file, it may be unsupported or corrupted",
     "ffmpeg.out_of_memory": "not enough available memory, can't continue",