Two bugs caused campaign briefing videos to play without audio on
Linux/macOS (OpenAL builds):

1. OpenALAudioStream::update() only restarted playback on AL_STOPPED.
   After reset() on a freshly created stream, alSourceStop() is a
   no-op on AL_INITIAL, so the source stays AL_INITIAL. Audio frames
   arriving via onFrame() called update(), but AL_INITIAL != AL_STOPPED
   skipped play() entirely.
   Fix: also restart on AL_INITIAL state.

2. LoadScreen::init() runs a blocking video loop that never reaches
   the main game loop, so TheVideoPlayer->update() (which calls
   FFmpegVideoStream::update() -> play() if not already playing) was
   never invoked. Audio source underruns had no recovery path.
   Fix: call TheVideoPlayer->update() after each frameNext().

Related: #38

… loop

Same blocking loop issue as SinglePlayerLoadScreen: the ChallengeLoadScreen
video loop never returns to the main game loop, so TheVideoPlayer->update()
was never called. Without it the OpenAL source could stay AL_INITIAL or
underrun with no recovery.

Related: #38

OpenAL-Soft on macOS auto-pauses all sources via Core Audio session
suspension when the application loses window focus. A momentary focus
loss during the shell-map to loading-screen transition caused the video
audio source to enter AL_PAUSED and never recover, resulting in silent
video playback for the rest of the session.

update() in OpenALAudioStream now handles AL_PAUSED alongside AL_STOPPED
and AL_INITIAL: whenever the source is not playing and buffers are queued,
alSourcePlay() is called to resume.

Also removes temporary diagnostic fprintfs that were used to trace the
root cause via logs/issue_38.log.

Issue: #38

…essions)

Keep the shared OpenAL stream from unqueueing processed buffers before restart, which could drop freshly queued briefing audio during video transitions.

Backport the stable FFmpeg-side fixes into the shared Generals path by forcing stream gain to 1.0f, rejecting invalid decoded frame sizes, and converting float FFmpeg audio to PCM16 before queueing into OpenAL.

Also remove the temporary audit instrumentation used to isolate the issue and update the April dev diary.

Fixes #38