Improve span2html parsing logic for inline elements

[khaykov]

Currently, this html <b>bold <i>italic</i> bold</b> after being parsed to span and back to html will become <b>bold </b><b><i>italic</i></b><b> bold</b>. This is not a bug, this is how Google originally implemented the parsing inside Html.java. It's based on nextSpanTransition method, which is simple and works just fine for Google, because the look of styled text does not change, and the resulting html is not really important for most people using the Html class. We care about html, so we need to change this at some point.