Figure composition

[darobin]

We don't currently detail how we could do figure composition: when one figure actually contains subfigures. This is generally useful, and also relevant to #52.

[petermr]

Here's the first example I came across. http://www.trialsjournal.com/content/16/1/354/figure/F5 It's actually one image, but logically it should be three. This sort of thing is very common. If this was a figure with 3 subfigures (each with their own caption) it's much easier to parse them into meaningful fighunks. (Of course authors are now "trained" to merge figures, but we should give them the opportunity to do it better. And separate figures could be easier for the journal to typeset