libobs: Fix audio-only output did not receive raw_audio

[norihiro]

Description

Fix audio-only output did not receive raw_audio callback at all.

Motivation and Context

I want to make a plugin that outputs mixed audio track(s).

When defining obs_output_info as below, the callback raw_audio was never called.

struct obs_output_info vban_output_info = {
        .id = ID_PREFIX "output",
        .flags = OBS_OUTPUT_AUDIO,
        .get_name = vban_out_get_name,
        .create = vban_out_create,
        .destroy = vban_out_destroy,
        .start = vban_out_start,
        .stop = vban_out_stop,
        .raw_audio = vban_out_raw_audio,
        .update = vban_out_update,
        .get_defaults = vban_out_get_defaults,
        .get_properties = vban_out_get_properties,
};

As a workaround, audio arrives if adding OBS_OUTPUT_VIDEO to flags and defining raw_video. However, I don't think this workaround is not ideal as it causes raw-video to be sent from the GPU.

How Has This Been Tested?

• Build norihiro/obs-vban/pull/1 with modification of vban_output_info as above code.
• Start the output.
• Check raw_audio is called.

OS: Fedora 35.

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• [x] My code has been run through clang-format.
• [x] I have read the contributing document.
• [x] My code is not on the master branch.
• [x] The code has been tested.
• [x] All commit messages are properly formatted and commits squashed where appropriate.
• [x] I have included updates to all appropriate documentation.

[norihiro]

Previously I fixed audio-only output using ffmpeg-output.

• https://github.com/obsproject/obs-studio/pull/5418

The output ffmpeg-output receives video frames because a flag OBS_OUTPUT_VIDEO is set and a callback raw_video is defined. The video frames are discarded inside ffmpeg-output.

This PR addresses a fix for output whose flag has only OBS_OUTPUT_AUDIO so that this is a different issue.

[norihiro]

Just rebased to the latest master branch.

[PatTheMav]

@Lain-B Any objections to this? Seems to only return only from the prepare_audio function if the source is audio-only, as no alignment with video timestamps is necessary.