In the PDF Viewer, we currently have metrics for loaded PDFs such as the page count, and PDF features used. We have no idea how big loaded PDFs are in terms of file size. Understanding the distribution of PDF file sizes will help answer questions like:

1) What are good values for |kDefaultRequestSize| and |kReadBufferSize| in chrome_pdf::DocumentLoader? Can we possibly better tune them for real world usage?

2) The entire PDF content is loaded into memory. How much memory is the PDF Viewer using just to store the downloaded PDF?

3) How often does the PDF Viewer deal with very large (e.g. 1 GB) PDFs? How much effort should we spend to optimize for this?