Update <use> trees when the referenced element is changed.

Update <use> trees when the referenced element is changed.

While trying to make <use> shadow tree creation always async I stumbled
on this bug where removing the element that the <use> references would
not clean up the ShadowRoot so if you removed the shape it was copying,
or swapped it with a new one, the <use> would continue to render the
original shape.

I also removed a synchronous render tree update whenever the instance
trees were invalidated. This synchronous update doesn't make any sense
in the modern world and it meant that if you had a complex shape you were 
referencing with a <use> and then mutated 5 attributes on it we would do 5
synchronous style updates!

For reference the synchronous style code traces back 4 years to here:
http://trac.webkit.org/changeset/58960

R=pdr@chromium.org,fmalita@chromium.org
BUG=380592

[f(malita), 2014-04-11 13:49:40]

lgtm

[pdr., 2014-04-11 13:54:24]

We've been slowly unwinding these inDocument checks from SVGUseElement and I
think this is another unneeded one. Nice! Assuming the tests pass, LGTM.

[esprehn, 2014-04-11 16:13:43]

On 2014/04/11 at 13:54:24, pdr wrote:
> We've been slowly unwinding these inDocument checks from SVGUseElement and I
think this is another unneeded one. Nice! Assuming the tests pass, LGTM.

It seems I cause ASSERTs, I'll need to figure out what's going on:

crash log for renderer (pid 14275):
STDOUT: <empty>
STDERR: ASSERTION FAILED: (*it)->shadowTreeElement()->correspondingElement()
STDERR: ../../third_party/WebKit/Source/core/svg/SVGElementInstance.cpp(188) :
static void
WebCore::SVGElementInstance::invalidateAllInstancesOfElement(WebCore::SVGElement
*)
STDERR: 1   0x5fba622e
WebCore::SVGElementInstance::invalidateAllInstancesOfElement(WebCore::SVGElement*)
STDERR: 2   0x5fc86e4d
WebCore::SVGElementInstance::InvalidationGuard::~InvalidationGuard()
STDERR: 3   0x5fc817db
WebCore::SVGElementInstance::InvalidationGuard::~InvalidationGuard()
STDERR: 4   0x5fca3065
WebCore::SVGUseElement::svgAttributeChanged(WebCore::QualifiedName const&)
STDERR: 5   0x5fb93440
WebCore::SVGElement::attributeChanged(WebCore::QualifiedName const&,
WTF::AtomicString const&, WebCore::Element::AttributeModificationReason)
STDERR: 6   0x5e7d7640 WebCore::Element::didAddAttribute(WebCore::QualifiedName
const&, WTF::AtomicString const&)
STDERR: 7   0x5e7d75b5
WebCore::Element::addAttributeInternal(WebCore::QualifiedName const&,
WTF::AtomicString const&, WebCore::Element::SynchronizationOfLazyAttribute)
STDERR: 8   0x5e7e0691 WebCore::Element::setAttributeInternal(unsigned long,
WebCore::QualifiedName const&, WTF::AtomicString const&,
WebCore::Element::SynchronizationOfLazyAttribute)
STDERR: 9   0x5e7cffb3 WebCore::Element::setAttribute(WTF::AtomicString const&,
WTF::AtomicString const&, WebCore::ExceptionState&)
STDERR: 10  0x5d71ea4d
WebCore::ElementV8Internal::setAttributeMethod(v8::FunctionCallbackInfo<v8::Value>
const&)
STDERR: 11  0x5d71068a
WebCore::ElementV8Internal::setAttributeMethodCallback(v8::FunctionCallbackInfo<v8::Value>
const&)
STDERR: 12  0x5b421a90 v8::internal::FunctionCallbackArguments::Call(void
(*)(v8::FunctionCallbackInfo<v8::Value> const&))
STDERR: 13  0x5b477ebe v8::internal::Builtin_HandleApiCall(int,
v8::internal::Object**, v8::internal::Isolate*)

[fs, 2014-06-04 12:47:51]

This fix now has a crbug to go with it - crbug.com/380592

[esprehn, 2016-02-10 00:24:51]

Someone should fix this someday. :)

[pdr., 2016-02-10 00:43:58]

On 2016/02/10 at 00:24:51, esprehn wrote:
> Someone should fix this someday. :)

Is this still a bug? SVGElementInstance no longer exists.

[esprehn, 2016-02-10 01:04:18]

On 2016/02/10 at 00:43:58, pdr wrote:
> On 2016/02/10 at 00:24:51, esprehn wrote:
> > Someone should fix this someday. :)
> 
> Is this still a bug? SVGElementInstance no longer exists.

I think the bug is around when to re-stamp the shadow dom not about the <use>
instance tree.

[pdr., 2016-02-10 05:15:22]

On 2016/02/10 at 01:04:18, esprehn wrote:
> On 2016/02/10 at 00:43:58, pdr wrote:
> > On 2016/02/10 at 00:24:51, esprehn wrote:
> > > Someone should fix this someday. :)
> > 
> > Is this still a bug? SVGElementInstance no longer exists.
> 
> I think the bug is around when to re-stamp the shadow dom not about the <use>
instance tree.

I got this in https://codereview.chromium.org/1498323002

Unfortunately, I had long forgotten about this patch :(

[fs, 2016-02-10 11:43:39]

On 2016/02/10 at 05:15:22, pdr wrote:
> On 2016/02/10 at 01:04:18, esprehn wrote:
> > On 2016/02/10 at 00:43:58, pdr wrote:
> > > On 2016/02/10 at 00:24:51, esprehn wrote:
> > > > Someone should fix this someday. :)
> > > 
> > > Is this still a bug? SVGElementInstance no longer exists.
> > 
> > I think the bug is around when to re-stamp the shadow dom not about the
<use> instance tree.
> 
> I got this in https://codereview.chromium.org/1498323002

And before that, https://codereview.chromium.org/1087653004 fixed an issue
similar to this with a somewhat similar fix (the removal of the inDocument()
check.)