Split graphemes and grapheme_indices into two methods

[KamilaBorowska]

s.extended_graphemes() is more readable than s.graphemes(true), as you don't have to think what does true mean here. Those methods were implemented as default methods in order to preserve backward compatibility if somebody implemented UnicodeSegmentation for their own types.

This is a feature, so second number in semantic version should be increased.

[Manishearth]

i kinda think making it extended_graphemes() will make people use it less. (perhaps make the false case legacy_graphemes?)

[tapeinosyne]

@Manishearth
i kinda think making it extended_graphemes() will make people use it less. (perhaps make the false case legacy_graphemes?)

A preference for legacy_graphemes is apparent in #22 as well.

[KamilaBorowska]

Oh, there was already an issue about this in issue tracker. Somehow missed that.

Unfortunately it's not really possible to have optional parameters in Rust, so changing graphemes(true) into graphemes() would be a breaking change. Definitely a good choice for unicode-segmentation 2.0.0 however.

But now that I think about it... I think it's feasible to have a trait with a different name which exports graphemes() method. This could be however confusing and UnicodeSegmentation is a good name, so it's probably not a good choice.