Extract path info from requests without decoding

The fix for Guice issue #745[1] causes getPathInfo() within the
GuiceFilter to return decoded values, eliminating the difference
between "foo/bar" and "foo%2Fbar". This is in spec with the servlet
standard, whose javadoc for getPathInfo[2] states that the return
value be "decoded by the web container".

Work around this by extracting the path part directly from the request
URI, which is unmodified by the container. This is copying the Guice
behavior prior to the bugfix. We do this with a static method rather
than using our own HttpServletRequestWrapper as we don't want to
produce a request wrapper that is not in spec.

[1] https://github.com/google/guice/issues/745
[2] http://docs.oracle.com/javaee/7/api/javax/servlet/http/HttpServletRequest.html#getPathInfo()

Change-Id: I48fb32710d45f573fb167c6344ad76541672858c

gerrit-httpd/src/main/java/com/google/gerrit/httpd/restapi/RestApiServlet.java

@@ -74,6 +74,7 @@ import com.google.gerrit.extensions.restapi.RestResource;
 import com.google.gerrit.extensions.restapi.RestView;
 import com.google.gerrit.extensions.restapi.TopLevelResource;
 import com.google.gerrit.extensions.restapi.UnprocessableEntityException;
+import com.google.gerrit.httpd.RequestUtil;
 import com.google.gerrit.httpd.WebSession;
 import com.google.gerrit.server.AccessPath;
 import com.google.gerrit.server.AnonymousUser;
@@ -871,7 +872,7 @@ public class RestApiServlet extends HttpServlet {
   }
 
   private static List<IdString> splitPath(HttpServletRequest req) {
-    String path = req.getPathInfo();
+    String path = RequestUtil.getEncodedPathInfo(req);
     if (Strings.isNullOrEmpty(path)) {
       return Collections.emptyList();
     }