Fix XHR argument parsing and synchronous event callbacks.

Derived from James Keane's patches
<https://github.com/google/caja/pull/1965> -- thanks to him for spotting
the problem. Issues in our tests and in onreadystatechange were spotted
in testing and additionally fixed.

Allegedly fixes <https://github.com/google/caja/issues/1878>.

* Pass the correct number of arguments to native XMLHttpRequest, rather
  than one too many. Among other things, this caused us to incorrectly
  default to synchronous XHR.
* Fixed onreadystatechange firing twice after synchronous XHR on Chrome
  due to using an unsound browser test rather than a feature test to
  decide whether to apply a workaround.
* Fixed lack of argument coercion for some uses of the async parameter.
  (Should have no observable effects, so this is just on general
  code structure principles.)

Supporting changes:
* Fixed test-cajajs-invocation testBuilderApiNetNoFetch accidentally
  depending on that synchronous XHR.
* Increased timeout for preliminary-meta from 10 to 30 seconds as
  I have experienced timeouts on that test only frequently, and it is
  inherently slow compared to others.

@ a83ec15cf5b13ffbe482cfb31830b7eb5490ccd9

[MarkM, 2015-07-08 22:12:51]

LGTM

https://codereview.appspot.com/251030043/diff/1/src/com/google/caja/plugin/do...
File src/com/google/caja/plugin/domado.js (right):

https://codereview.appspot.com/251030043/diff/1/src/com/google/caja/plugin/do...
src/com/google/caja/plugin/domado.js:1052: switch (arguments.length - 1) {
Is this really clearer than switching on argument.length and incrementing the
case labels? If not, please do that.

[kpreid_google, 2015-07-09 17:47:11]

https://codereview.appspot.com/251030043/diff/1/src/com/google/caja/plugin/do...
File src/com/google/caja/plugin/domado.js (right):

https://codereview.appspot.com/251030043/diff/1/src/com/google/caja/plugin/do...
src/com/google/caja/plugin/domado.js:1052: switch (arguments.length - 1) {
On 2015/07/08 22:12:51, MarkM wrote:
> Is this really clearer than switching on argument.length and incrementing the
> case labels? If not, please do that.

I think that the numbers currently used are more relevant to _what we are
implementing_ (XHR API) as opposed to the +1 numbers which contain an
implementation detail (the privates arg).

Do you disagree?

[MarkM, 2015-07-09 22:25:37]

On 2015/07/09 17:47:11, kpreid_google wrote:
>
https://codereview.appspot.com/251030043/diff/1/src/com/google/caja/plugin/do...
> File src/com/google/caja/plugin/domado.js (right):
> 
>
https://codereview.appspot.com/251030043/diff/1/src/com/google/caja/plugin/do...
> src/com/google/caja/plugin/domado.js:1052: switch (arguments.length - 1) {
> On 2015/07/08 22:12:51, MarkM wrote:
> > Is this really clearer than switching on argument.length and incrementing
the
> > case labels? If not, please do that.
> 
> I think that the numbers currently used are more relevant to _what we are
> implementing_ (XHR API) as opposed to the +1 numbers which contain an
> implementation detail (the privates arg).
> 
> Do you disagree?

Yes, for this case I agree. Still LGTM