Reverse of wrapUntrustedContent. Extracts the original body text
(un-escaping any sentinel-encoded close tags). Throws if the input is not
a single well-formed fence — callers shouldn't be parsing arbitrary
markdown with this; it's for tests + symmetric reasoning only.
Reverse of wrapUntrustedContent. Extracts the original body text (un-escaping any sentinel-encoded close tags). Throws if the input is not a single well-formed fence — callers shouldn't be parsing arbitrary markdown with this; it's for tests + symmetric reasoning only.