I created a patch based on Sergey Ushakov's work that handles multipage TIFFs. This allows fast and efficient conversion from TIFF to PDF
The general approach is to provide a new factory method that accepts an image (page) number, and then appropriate page number is located when the CCITT stream is being extracted.
There's a minor inefficiency in this approach because the seek starts from the beginning for each page, causing O(N^2) algorithm when extracting every page, but maximum size for file appears to be 2 GB and the cost for finding a single page will still be low, so I bet this will never come up in practice.
There is no method that tells how many pages TIFF files have. I opted to simply return null from the factory method that accepts page number if there is no such page, so users can use this as condition to break from a TIFF to PDF conversion loop.