Render AST/model directly to iOS AttributedString

[stefanceriu]

To improve performance by reducing the amount of data mapping. To improve maintainability by removing the need to map selection indices between the different model formats (AST / HTML / AttributedString)

Would fix element-hq/element-x-ios#2030 and element-hq/element-ios#7489

[Velin92]

Btw another easier fix for the meantime to help with element-hq/element-x-ios#2030 and element-hq/element-ios#7489 would be to try the following solution found by @stefanceriu to intercept that a text to speech is happening:
https://stackoverflow.com/questions/32652775/ios-how-to-detect-if-voice-dictation-was-used-for-uitextfield-or-microphone-bu/53440078#53440078

Maybe we could use this to avoid using the diffing system until we are sure that the user has ended the dictation?