SDL3 example '01-simple-playback': minimum_audio is calculated with input format instead of output format

[nathanvdh]

I think this probably also applies to 03-load-wav and 04-multiple-streams.
SDL_GetAudioStreamAvailable() returns

the number of converted/resampled bytes available

that is: the number of bytes after resampling the input to destination/device format.

I expect this is typically not the same as the the generated input 8000hz, 32-bit float mono input. E.g. on my Ubuntu laptop with headphones plugged in its 44100Hz, S16, 2 channel. This example does a comparison that is not apples-to-apples:

    const int minimum_audio = (8000 * sizeof (float)) / 2;  /* 8000 float samples per second. Half of that. */
    if (SDL_GetAudioStreamAvailable(stream) < minimum_audio) {
    ...

I discovered this while doing some network streaming and playback, I kept getting data underflow and injected silence, because I thought my buffers were satisfied when they were not. minimum_audio should be calculated using the device format gotten via a SDL_GetAudioStreamFormat(stream, NULL, &dst_spec).

Edit: I've also just seen SDL_GetAudioStreamQueued(), which returns the amount of queued input data, so that could be used too, instead of SDL_GetAudioStreamAvailable().

[icculus]

I was just coming in to say the real bug is we should be using SDL_GetAudioStreamQueued, but I see from your edit that you realized this, too. :)

I'm fixing this up real quick.