load_score based on file extension

[fosfrancesco]

Right now load_score works by trying all possible parsers in try-except. While this seems nice, so it can handle files with slightly different extensions strings, it creates many other problems:

• speed: for file types at the end of the try except list, all the other parsers need to start and fail before the right one is used
• failing clause: sometimes a parser just doesn't fail and yields invalid results. For example there is not universal way of knowing for example if a file is of XML type. I implemented some heuristics for the Kern and MEI parser to stop them earlier, but I'm not sure this is covering all possible cases.

For example I was trying to load musescore scores without having musescore installed. Instead of an error it was loading the musescore file as match file. Producing a score with 0 parts.

All considered I vote to use an extension-based system, where a set of possible extension strings is hardcoded for each file type. E.g., for XML ".musicxml, .mxml, .mxl, .MusicXML, etc.."

[gregchapman-dev]

From my experience at Apple, where the same software (QuickTime first, and then AVFoundation) could open anything from .mov to .mp3 to .jpg: we insisted on only using file extension, and never sniffing inside the file to see which parser to use. And we were VERY glad of that decision over the years.

[neosatrapahereje]

I guess this could be a good change, which will reduce some overhead of the try except clauses. Although I was in favor of the original implementation, I've come to see that there are a few quirks that would require unnecessarily convoluted bolierplate code. So yes, I think it would be better to just check the extension.

[fosfrancesco]

Ok, I will work on that then!