Refactor metadata extraction

[ralsina]

Parts of it can be refactored into format-specific extractors (in the compiler plugins) which would get it from places like the docutils document metadata.

[damianavila]

This is interesting for ipynb too... notebooks can have metadata where we could post meta info...

[ralsina]

One concern is that to get the metadata we need to parse the thing. So, we may need to cache the intermediate representation (like the doctree in rst)

[Kwpolska]

We already have read_metadata() (#1603), used by eg. ipynb: https://github.com/getnikola/nikola/blob/master/nikola/plugins/compile/ipynb.py#L125

Its performance is not perfect (we read files twice), but that’s fixable too if it’s a concern.

[ralsina]

Ok, so if I implemented read_metadata in, say, the reSt compiler, that would do things. I may take a shot at it.

[Kwpolska]

For #1923, perhaps? (feel free to take it over, I didn’t get much stuff done for that)

[ralsina]

Yes, that too. We could even use yaml metadata or whatever.

[ralsina]

There are now metadata extractors in the reSt, markdown and ipynb compilers. I say, let's close.