Bug 1560055 part 3. Switch various "width" and "height" attributes on HTML elements to mostly follow the spec for parsing the attribute. r=mccr8

The spec allows non-integer values, but we don't have a good way to store them
in nsAttrValue yet.  See https://bugzilla.mozilla.org/show_bug.cgi?id=1561440

HTMLTableCellElement::MapAttributesIntoRule can now call
MapImageSizeAttributesInto instead of manually mapping width and height, because
0 values (which it was excluding before) are now excluded at attribute parse
time.

For 'width' on HTMLTableElement I kept our old behavior for 0, which matches the spec
but not Safari or Chrome.

For 'height' on HTMLTableElement I kept our old behavior for 0, which matches
Safari and Chrome but not the spec.  https://github.com/whatwg/html/issues/4715
tracks a possible spec change.

Same thing for 'height' on HTMLTableRowElement.

Same thing for 'width' on HTMLTableColElement.

The ParseImageAttribute call in HTMLMediaElement is not needed, because
HTMLAudioElement does not map any of those to style and HTMLVideoElement only
maps width/height, which it already parses.

Differential Revision: https://phabricator.services.mozilla.com/D36127

dom/html/HTMLInputElement.cpp

@@ -5185,10 +5185,10 @@ bool HTMLInputElement::ParseAttribute(int32_t aNamespaceID, nsAtom* aAttribute,
       return true;
     }
     if (aAttribute == nsGkAtoms::width) {
-      return aResult.ParseSpecialIntValue(aValue);
+      return aResult.ParseHTMLDimension(aValue);
     }
     if (aAttribute == nsGkAtoms::height) {
-      return aResult.ParseSpecialIntValue(aValue);
+      return aResult.ParseHTMLDimension(aValue);
     }
     if (aAttribute == nsGkAtoms::maxlength) {
       return aResult.ParseNonNegativeIntValue(aValue);