Factor out common code between pdf and ps backends.

[anntzer]

(The AFM cache only needs to cache on filename, not on font properties,
because the fontprop->filename conversion is already cached by
font_manager.)

PR Summary

PR Checklist

• Has Pytest style unit tests
• Code is PEP 8 compliant
• New features are documented, with examples if plot related
• Documentation is sphinx and numpydoc compliant
• Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

[jklymak]

Something is wrong w/ py 2.7 figures.

[anntzer]

forgot to future-import division :p

[jklymak]

appveyor seems to have picked up a bug. Both your recent PRs have it....

[anntzer]

Seems unrelated, tbh...

[jklymak]

No I agree, it has nothing to do w/ your PRs, but I suspect a bug was recently introduced into Master

[QuLogic]

The last build to pass had pytest 3.2.5; this one has 3.3.0. There was a similar bug in 3.2.1: pytest-dev/pytest#2644 but maybe there was a regression.

[QuLogic]

New report is at pytest-dev/pytest#2957.

[jklymak]

This seems like a useful re-factor. It could use a second review... ping @tacaswell @jkseppan

[anntzer]

rebased

Additionally merged the implementations of get_canvas_width_height, with the following addition to the commit message:

get_canvas_width_height can be merged because FigureCanvasPdf always
makes sure to call newPage(width, height) and create a new RendererPdf
with the same width and height, and it's likely other things break if
the PdfFile and the Renderer's size don't match anyways.

[anntzer]

@jklymak @timhoffm either of you want to click on the green button? :)

[anntzer]

thanks

[jklymak]

Thank you!