Bold/Emphasis regression

I apologize if this is a repeat --- I thought I emailed you about this, but now cannot find that email. So perhaps I never sent it.

A while back you fixed a problem with the strong/emph parsing that could cause the program to either fail to parse or to take a long time to parse. That worked great, however it caused a regression where the following no longer parses correctly (example sent by one of my users):

```
The ***quick*** brown ***fox*** jumped
```

The problem appears to be introduced in the 2a19a38d075356650c59206069fa925028182522 commit, and remains in the latest commit of peg-markdown.

Thanks!

Fletcher