Further bindings idea: Mozilla pdf.js with JSPyBridge and node-canvas

[mara004]

As in #230, yet another python pdf bindings idea, see the title.

I already tried this a year ago or so and was able to perform some non-rendering tasks like getting PDF version, but got stuck being unable to acquire a canvas object, since the special document property is only defined in a browser.
However, I now found node-canvas, which appears to be a non-browser re-implementation of Mozilla's web canvas API, so it looks like rendering with pdf.js from Python might be doable nevertheless.

[mara004]

Draft: https://gist.github.com/mara004/87276da4f8be31c80c38036c6ab667d7

[mara004]

Unfortunately, it turns out buffer transfer is very inefficient with JSPyBridge ATM, but we can revisit the situation if extremeheat/JSPyBridge#67 (comment) gets addressed.
In general, the implementation of JSPyBridge doesn't seem elegant. It doesn't interact with javascript at VM/interpreter level, but merely communicates in a pipe-like fashion (currently abusing stderr), i.e. no direct integration, unlike ctypes (->C) or jpype (->Java).

[mara004]

Transfer speeded up via extremeheat/JSPyBridge#103 (new API blobValueOf()).