CSS: Don't override font-weight for section

.slides>section and .slides>section>section set the font-weight to
"normal". This overrides any font-weight setting a theme may set at
the body or .reveal level, and requires a theme author to also
specifically set the font-weight for .slides>section and
.slides>section>section. That's tedious and also counterintuitive.

Adopt a saner default by setting the font-weight to inherit.

It should be `{. xxxx=yyy }`, not `{_ xxxx=yyy }`.
Only test/test-element-attributes-markdown.html were using `{_`,
simply in order to test the ability, for the user, to define a _different_ pattern.

It is important that test/test-element-attributes-markdown.html do test
the ability for the user to define its own pattern.

The first slide of a vertical stack see some data attributes ignored.

Mainly the data-transition one.

Repeat all data-attributes on the wrapping section element.

Ignore any other attributes (like 'id="xxx"'), in order to not
mess with internal links (by repeating twice an id).

By default, look for <!-- slide-attributes: xxxx -->.
Whatever 'xxx' is will be added to the section attributes.

You can define your own pattern with 'data-attributes'.
For instance 'data-attributes="^\s*?-- (.*?)$"':
that will be added to the options.

The 'attributes' section is removed from the content of the slide,
so the final markdown doesn't reflect them.
That also means you can add those attributes *anywhere* in the slide

But that allows for *any* attribute to be added for a specifc slide,
like:

- id="plan", for allowing internal link (issue #430)
- data-background="#ff0000"
- data-transition="fade"

You list those attributes on a single line, like

- (default):
` <!-- id="plan" data-background="#ff0000" -->`
- or, with an alternative data-attributes pattern:
` -- id="plan" data-background="#ff0000"`

Again, that line is remove from the final content.

Test paragraphs, list items and img attributes.

By default '{\\\.\s*?([^}]+?)}'.

Allows to add attributes to element of an attribute which contains
the attribute pattern.

Works better, and the html elements get their attributes.

Extra text representing classes is detected and correctly removed.
Adding attributes isn't working yet.