Idea: track prebuffering, possibly with configurable buffer size

[kFYatek]

Feature Description

When loading tracks from unreliable sources, like network shares (for some reason this issue hits much worse on macOS than on Windows), the audio can drop if the file access is not instant enough.

Modern computers have plenty of memory to buffer (including decoding into PCM in case of compressed files) entire tracks into RAM as they are loaded into decks in the UI. That would allow the audio to be loaded and decoded asynchronously, without relying on any performance guarantees regarding file access.

In case of running on memory-constrained systems and/or dealing with exceptionally long tracks, having configurable buffer would allow to limit how much of the track is prebuffered. It would still allow loading the missing parts asynchronously, several minutes before they are actually going to be played.

All that would make playing from things like NAS devices more reliable and practical.