'<' is not encoded when convert markdown to html

[cnblogs-dudu]

Use command line: pandoc -f markdown_github -t html
Input markdown text: https://gist.github.com/cnblogs-dudu/c9fde76518998e5386f6e66bd89b281a

Get the following html when use pandoc 1.17.2 or pandoc 1.18.

<p>i（i<j)

```javascript
function binarySearch1(a, target) {
  while(start <= end) {
    var mid = ~~((start + end) >&gt; 1);<br />
}<br />
}<br />
```</p>

Expect the following right html. It works on pandoc 1.17.

<p>i（i&lt;j)</p>
<div class="sourceCode"><pre class="sourceCode javascript"><code class="sourceCode javascript"><span class="kw">function</span> <span class="at">binarySearch1</span>(a<span class="op">,</span> target) <span class="op">{</span>
  <span class="cf">while</span>(start <span class="op">&lt;=</span> end) <span class="op">{</span>
    <span class="kw">var</span> mid <span class="op">=</span> <span class="op">~~</span>((start <span class="op">+</span> end) <span class="op">&gt;&gt;</span> <span class="dv">1</span>)<span class="op">;</span>    
  <span class="op">}</span>
<span class="op">}</span></code></pre></div>

[jgm]

I'm pretty sure this is because of commit 773bbb8 We used to fail if the HTML tokenizer yielded any warnings. This commit changed this...in fact we got lots of complaints about pandoc's failure to parse malformed HTML. Unfortunately the same change that allows pandoc's HTML reader to accept things like unescaped `&` also allow your case. So this needs more thought.

[jgm]

Note: I recommend that if you are including math, you either backslash-escape the `<` sign or, better, use tex math. i\<j and j\>k or $i<j \text{and} j>k$ Otherwise there's always a risk of interpretation as HTML, and this would remain even if we made the parser stricter. E.g. in x<a and y>z we parse as an "a" element with boolean attributes "and" and "y". (Yes, those aren't official HTML attributes, but we don't validate tags and attributes, because we want pandoc to be flexible enough to accept custom elements, docbook, etc.)