Replace wkhtmltopdf with something else

[benjaminwood]

While it's nice that wkhtmltopdf is a single executable dependency, it isn't reliable. For example, there is no way to define a max wait time when loading an image. This means that conversions can fail if images don't load in time.

Other options may include:

• Headless chrome
• Conversion as a service: cloudconvert.com
• athenapdf (can't do this on lambda, though): https://github.com/arachnys/athenapdf