Stop decoding all %2e's in paths

This basically reverts bee5ad8 as it
did not end up shipping in Firefox and the tentative conclusion is that
percent-decoding in URLs while parsing creates more trouble than it’s
worth.

We’ll still handle %2e when it’s part of a single-dot or double-dot
path segment.

Tests: web-platform-tests/wpt#4104.

Closes #87.

url.bs

@@ -1900,12 +1900,8 @@ string <var>input</var>, optionally with a <a>base URL</a> <var>base</var>, opti
        <li><p>If <a>c</a> is "<code>%</code>" and <a>remaining</a> does
        not start with two <a>ASCII hex digits</a>, <a>syntax violation</a>.
 
-       <li><p>If <a>c</a> is "<code>%</code>" and <a>remaining</a>, <a>ASCII lowercased</a> starts
-       with "<code>2e</code>", append "<code>.</code>" to <var>buffer</var> and increase
-       <var>pointer</var> by two.
-
-       <li><p>Otherwise, <a>UTF-8 percent encode</a> <a>c</a> using the <a>default encode set</a>,
-       and append the result to <var>buffer</var>.
+       <li><p><a>UTF-8 percent encode</a> <a>c</a> using the <a>default encode set</a>, and append
+       the result to <var>buffer</var>.
       </ol>
     </ol>