In richtext2, do not require special handling of class="Apple-style-span"

[GoogleCodeExporter]

The tests RTE2-C_BC:ace_FONT.ass.s:bc:rgb-1_SW and 
RTE2-C_HC:ace_FONT.ass.s:bc:rgb-1_SW both test that if changing the background 
color on an input like

  <span class="Apple-style-span" style="background-color: rgb(255, 0, 0)">[foobarbaz]</span>

the result is something like

  <span style="background-color: #aaccee">[foobarbaz]</span>

However, the algorithm in the editing spec, and as of 
<https://bugzilla.mozilla.org/show_bug.cgi?id=746515#c8> also the algorithm in 
Gecko, works by first removing all styles, then adding new styles in a wrapper 
span, without adding styles to any existing elements.  This makes the behavior 
in CSS mode match the behavior in HTML mode, where as a rule, only new elements 
are added and existing ones aren't modified.  (Modifying existing elements can 
make the modes behave differently in some cases when properties interact, e.g., 
color and text-decoration.)  Thus the styles are first stripped, leaving

  <span class="Apple-style-span">[foobarbaz]</span>

(because it won't remove the span because it has a class), and then a new span 
is added, giving

  <span style="background-color: rgb(170, 204, 238);"><span class="Apple-style-span">[foobarbaz]</span></span>

This has an extra <span>, so it's not as clean.  But the alternatives are either

1) Special-case class="Apple-style-span".  This is not acceptable in a 
standard.  WebKit no longer produces it anyway, so it shouldn't be necessary 
going forward.

2) Add style="" to existing elements, not just newly-created ones.  This is not 
desirable because it can lead to CSS and non-CSS modes working differently, 
like in the case of execCommand("underline") on <span class="x">foo</span> with 
the style rule .x { color: red }.  If in CSS mode we produced <span class="x" 
style="text-decoration: underline">foo</span>, the underline would be red.  But 
in non-CSS mode we have to produce <u><span class="x">foo</span></u>, and the 
underline will be black.  To avoid this possibility, we need to avoid adding 
style="" to any existing elements.

Please update the tests to permit the result required by the spec.  Thank you.

Original issue reported on code.google.com by a...@aryeh.name on 20 Apr 2012 at 6:21