Combine "get visible client rect" functions into a common place

[peitschie]

A common UI task is translating a node/offset pair into a visible client rect (aka: determine the dimensions of a visual caret if it was to be drawn). Each implementation currently has it's own set of quirks and problems, and the inconsistencies make cross-browser work difficult (e.g., issue #224).

Implementations in the existing code include:
gui.Caret.getSelectionRect
gui.Caret.verticalOverlap (used to detect white-space word wrapping)
gui.SelectionMover.getVisibleRect
gui.SelectionView.firstVisibleRect
gui.SelectionView.lastVisibleRect