Use fetch instead of FileLoader in SplatLoader to avoid reusing same arraybuffer

[mrxz]

A somewhat crude workaround for #93. Instead of using the FileLoader it uses the Fetch API directly. Goal is to fix the regression of the envmap example. There are still some points that can be improved:

• The onProgress callback is no longer called
• The error could still occur with SOGS as the "extra files" are still loaded using the fileLoader
• Loading the same file/URL will always result in separate fetches (no caching or combining concurrent requests)

[asundqui]

Wow thanks for finding this issue and fixing it so quickly. I just PR and merged another addition to parse SOGS from within a .Zip archive, hopefully it doesn't collide with this. But I'd say let's use your fix! Agree we could improve things in the future... probably at some point the whole file loading etc needs a cleanup and revamp, lots of paths now as a result of different file format support.

[asundqui]

@mrxz would it be possible to make this handle onProgress callback? What would that involve, having it collect an array of ArrayBuffers for each packet, and then merging them at the end? I guess I can live with not handling all errors perfectly right now...