…ng issues, update getAsCssString

Add CSSNestedDeclarations class to represent nested declarations, like the
CSSOM does. CSSStyleRule now has a list of ICSSNestedRule, which includes
various rules such as CSSStyleRule or CSSMediaRule. See
https://drafts.csswg.org/css-nesting-1/#conditionals

Update CSSNodeToDomainObject to create CSSNestedDeclarations instances where needed.

Add support for parsing nested at-rules, e.g.: .foo { @media print {} }

Fix parsing issue where the parser confused an element selector with a style
declaration, e.g. such as in the following 2 cases:
* .foo { p { color: red } }
* .foo { p: value; }
Add a lookahead that checks if a style declaration follows, and only consume it
in that case. Otherwise, proceed to look for rules and nested declarations.

Add a new parser rule for consuming a style declaration with nested elements.
Not all rules support nesting, e.g. @font-face. For these rules, we do want
to keep the previous behavior where it throws on nested rules.

Minor change to how whitespace is handled: getAsCssString only outputs itself
as a string, it does not output any leading or trailing whitespace or newlines.
Space and lines around an CSS element should be handled by the container that
contains the element, as it may depend on the context of that container.

& is now a valid selector member. The test suite has changed the tests to use %
instead of &, which is still invalid. See:

https://github.com/web-platform-tests/wpt/blob/master/css/selectors/old-tests/css3-modsel-156.xml

support for relative selectors in nested style rules, e.g.

.foo { > .bar { ... } }

When a style rule is nested, it allows a <relativeSelector>, not just a
<selector>. But at the top-level, a <relativeSelector> is not allowed, i.e.
the following is invalid:

> .bar { ... }

To prevent the grammar from blowing up, always allow <relativeSelector> in a
style rule, even at the top-level. Check for disallowed <relativeSelector> at
the top level when mapping the parsed AST to the domain model. Issue a CSS
interpretation error if such an invalid relative selector is encountered.

The file is not actually good -- it's missing an rbrace!

…rule

With the browser compliant flag enabled, the behavior did not change (empty string).

With the browser compliant flag disabled,

.class{color:red;.class{color:green}.class{color:blue}

is now parsed as nothing. Previously, an error occurred at the second ".class", causing the
parser to skip to the next rbrace. Now, the second ".class" is the start of a nested
rule. The parser only encounters an error at the very end, when it finds a missing rbrace.
The error recovery tries looking for the next rbrace, but finds none, only an EOF, causing
it to discard the entire rule.

Use prefix "s" for string variables
Use prefix "m_" for instance fields

As per https://github.com/phax/meta/blob/master/CodingStyleguide.md