PDF export of a request

Part of #18: We should be able to build a PDF record of each request.

I imagine this will require:

1. Write an HTML page describing the request. Render this to a PDF (wkhtmltopdf? pdfire?)
2. Convert all image receipts to PDFs. (or include them in the HTML above)
3. Concatenate into a single PDF.

If we had receipt thumbnails, we could use them to refer to receipts (like expensify does).

Complexity: Dealing with untrusted PDFs and images in the same application as our data is not ideal. We should probably have a separate sandboxed worker for thumbnailling and building PDFs