Consolidate navigation-timing tests into web-platform-tests

LayoutTests/http/tests/w3c/webperf/approved/navigation-timing/ and
LayoutTests/external/wpt/navigation-timing/ have the same origin but
have diverged slightly over time.

https://chromium-review.googlesource.com/c/chromium/src/+/782419 show
what the interesting differences between the two copies were.

metadata_cache in most of the tests were removed when the tests were
initially added to web-platform-tests:
#181

idlharness.html: It's OK to no longer test an instance of Window,
that's being done in the hr-time test:
#8388

test_timing_xserver_redirect.html: The bits around pageOrigin were
removed in 234a672.
That assert was still failing in our own tests, but removing it is OK,
using location.hostname is wrong as it doesn't include the port:
https://html.spec.whatwg.org/multipage/links.html#dom-hyperlink-hostname

Change-Id: Ic5c4151a3e309fee1894c72cd14d159bd314be20

navigation-timing/idlharness.html

@@ -19,11 +19,6 @@ <h1>Navigation Timing IDL tests</h1>
 <pre id='untested_idl' style='display:none'>
 typedef double DOMHighResTimeStamp;
 
-interface Window {
-    [Replaceable]
-    readonly attribute Performance performance;
-};
-
 [Exposed=(Window,Worker)]
 interface Performance {
 };
@@ -140,7 +135,8 @@ <h1>Navigation Timing IDL tests</h1>
   idl_array.add_untested_idls(document.getElementById("untested_idl").textContent);
   idl_array.add_idls(document.getElementById("idl").textContent);
 
-  idl_array.add_objects({PerformanceNavigation: ["window.performance.navigation"],
+  idl_array.add_objects({Performance: ["window.performance"],
+                         PerformanceNavigation: ["window.performance.navigation"],
                          PerformanceTiming: ["window.performance.timing"],
                          PerformanceNavigationTiming: ["window.performance.getEntriesByType('navigation')[0]"] });