libobs: Apply resampler offset to system audio TS

Instead of applying the resampler offset right away (to each audio
packet), apply the resampler offset when the timestamps are converted to
system timestamps.  This fixes an issue where if audio timestamps reset
to 0 (for whatever reason), the offset would cause the timestamp to go
in to the negative.

libobs/obs-internal.h

@@ -528,6 +528,7 @@ struct obs_source {
 	/* timing (if video is present, is based upon video) */
 	volatile bool                   timing_set;
 	volatile uint64_t               timing_adjust;
+	uint64_t                        resample_offset;
 	uint64_t                        next_audio_ts_min;
 	uint64_t                        next_audio_sys_ts_min;
 	uint64_t                        last_frame_ts;

libobs/obs-source.c

@@ -1149,6 +1149,8 @@ static void source_output_audio_data(obs_source_t *source,
 			push_back = true;
 	}
 
+	in.timestamp -= source->resample_offset;
+
 	source->next_audio_sys_ts_min = source->next_audio_ts_min +
 		source->timing_adjust + source->sync_offset;
 
@@ -2219,6 +2221,7 @@ static inline void reset_resampler(obs_source_t *source,
 
 	audio_resampler_destroy(source->resampler);
 	source->resampler = NULL;
+	source->resample_offset = 0;
 
 	if (source->sample_info.samples_per_sec == obs_info->samples_per_sec &&
 	    source->sample_info.format          == obs_info->format          &&
@@ -2298,16 +2301,15 @@ static void process_audio(obs_source_t *source,
 
 	if (source->resampler) {
 		uint8_t  *output[MAX_AV_PLANES];
-		uint64_t offset;
 
 		memset(output, 0, sizeof(output));
 
 		audio_resampler_resample(source->resampler,
-				output, &frames, &offset,
+				output, &frames, &source->resample_offset,
 				audio->data, audio->frames);
 
 		copy_audio_data(source, (const uint8_t *const *)output, frames,
-				audio->timestamp - offset);
+				audio->timestamp);
 	} else {
 		copy_audio_data(source, audio->data, audio->frames,
 				audio->timestamp);