Improve handling of newline in inline code block

[mikeando]

This case `\n` was getting parsed as <code></code> rather than <code> </code> due to the \n getting flagged as a non-space character. Change is pretty simple

There's actually an unrelated commit with this that changes error reporting to use #[track_callers] that makes test failures point to the right line (rather than inside the compare_strings helper function)

I can break this into two PRs if you want -- but they seemed trivial enough to review together.

[mikeando]

Ugh, there is something trickier going on than I'd thought.
The same node ends up getting processed for inlines more than once.
It will be a day or so before I have time to dig through what is going on...
Changing this PR to WIP until then.

[mikeando]

The trouble I was having was actually with the round-trip tests, not where I thought it was.
They're fixed too now I think. And all the spec tests pass on my machine.

The cause ofmy confusion was that I didn't realise the html function was doing more than just
checking the output matched the provided HTML (and actually does an additional round-trip check).
I then added a bunch of debug prints that sent me down the wrong path altogether. Spent a long time
trying to understand why my Code node was getting processed twice... with different content.

Anyway, sorry about the delay on this.

[ghost]

ta, this looks superb! I ran cargo fmt to clear up the last check and pushed. It looks like this change mighta crept in between the last two CM spec versions — at some point we should actually make sure we're compliant with 0.30.

[ghost]

if you want a release with these changes, give us a shout, otherwise we'll just wait until there's more accumulated.

[mikeando]

No need for a new release. I'm happy to point my cargo at the commit for now.