Decode hash value before querying for named anchor

With a URL like `#%24.ajax`, we need to first decode the value before
searching the DOM for the element named `$.ajax`.

In Firefox, this is not necessary, as the browser already auto-decodes
hash values. Therefore, in Firefox this will perform double-decoding,
which could theoretically lead to inability to match anchors that
contain a percent-encoded value literally. However, since Firefox also
seems to automatically decode values for anchors named via `name="..."`
HTML attribute, you're going to have a bad time anyway with named
anchors that contain percent-encoded characters literally, with or
without pjax.

Author: mislav

jquery.pjax.js

@@ -327,7 +327,7 @@ function pjax(options) {
       pjax.state.url = url.href
       window.history.replaceState(pjax.state, container.title, url.href)
 
-      var name = hash.slice(1)
+      var name = decodeURIComponent(hash.slice(1))
       var target = document.getElementById(name) || document.getElementsByName(name)[0]
       if (target) $(window).scrollTop($(target).offset().top)
     }

test/unit/pjax_fallback.js

@@ -170,7 +170,23 @@ asyncTest("scrolls to anchor at bottom page"+s, function() {
   })
 })
 
+asyncTest("scrolls to named encoded anchor"+s, function() {
+  var frame = this.frame
+
+  equal(frame.window.scrollY, 0)
+
+  this.loaded = function(frame) {
+    setTimeout(function() {
+      equal(frame.window.scrollY, 10008)
+      start()
+    }, 10)
+  }
 
+  frame.$.pjax({
+    url: "/anchor.html#%62%6F%74%74%6F%6D",
+    container: "#main"
+  })
+})
 
 asyncTest("sets GET method"+s, function() {
   var frame = this.frame