Allow iter_all to iterate over multiple classes

[leleogere]

Fixes #426.

When iterating over a Part's elements, the user can only provide a single class to filter the output. If they want to get multiple classes, they can either iterate over all classes and filter a posteriori, or iterate multiple times over a single class. Those two workarounds are quite slow, so the possibility to iterate over multiple classes in a single pass would be useful.

This PR simply allows the user to pass an iterable to the cls parameter of iter_all. For example, the following would allow to easily retrieve a list of all key/time signature/clef change over the score (in that order when they are simultaneous), in a single iteration over all time points.

part.iter_all(cls=[TimeSignature, KeySignature, Clef])

There are still a couple of things to discuss before merging this, briefly evoked in #426:

• It makes sense that the parameters start, end and mode apply to all classes. However, should the parameter include_subclasses apply to all of them? Maybe I could add the possibility to pass either a single boolean (apply to all classes) or an iterable of booleans (apply to indivisual classes, same size as the cls iterable).
• What if a user provides cls=[GenericNote, Note] with include_subclasses=True? (Or even simpler if they provide cls=[Note, Note]) Should the iterator return duplicate objects? Or deduplicated objects only? Or raise an error?

I have not added any tests yet, but I can add some, although I'm not sure in which file they would belong. I haven't found any existing test for the function iter_all.