Wrong word cut at end of lines with spaces

Bug 5185017: when the line length is exceeded at a space character,
we use the previous ok width, and the last word is wrapped to next line
although it fits.

This back-track also generates problem with the span parsing, where the
spanStart indexes are no longer monotonuously increasing.

Plus some refactoring in this code (unused parameters, calls to out())

Change-Id: Ia8cd310a732752af3bd370bf0a16db23d40e83f2