Document migration from 0.x to 1.0

[masklinn]

It's not complicated but still...

Maybe should also explain the increase in depth of the model? Or is it more for #182? Whichever location it is, it should use write more classes as its basis for explanation. Also probably make sure I don't misunderstand and didn't misremember it.

important bits:

• new API is more lazy by default, global parser is only instantiated when first accessed, not when imported, and that's the moment at which the matchers are instantiated
• matchers can be made even lazier (by using lazy matchers), though that's probably not a great idea with the basic parser
• re2 exists
• most difficult part of the migration is yaml, as the implicit loading was replaced