Content parsed with InlineCodeFinder should be encoded on merge...

InlineCodeFinder can’t know if the content it is processing is a subformat (for example HTML) - but we should minimally encode the content for the parent format. Example TAML with HTML tags:

input:

<a href=\"/users/unlock/new\">request another one</a>

incorrect output:

<a href="/users/unlock/new">request another one</a>

The YAML filter decodes the \” - but does not restore them on merge. This may happen with other filters that use InlineCodeFinder. We should come up with a consistent way to handle this. Should we also handle subformats?

‌