feat: use pikepdf instead of PyPDF2, bumps Python min. version to 3.6

Several reasons:
1. PyPDF2 isn't maintained anymore.
2. On PDF files with lots of pages, you hit a recursion limit because of
   the way PyPDF2 is written. The `_sweepIndirectReferences` function is
   recursive instead of being iterative.
3. Performances. PyPDF2 is a pure Python library, while pikepdf uses
   QPDF, a C++ library under the hood. It is much faster. This is quite
   noticable when processing PDFs such as books.
4. pikepdf fixes PDF. While implementing the ToC feature, I noticed that
   for some PDFs it didn't show up on the Remarkable, even before
   processing it. For some reason, simply opening a PDF with pikepdf and
   saving it again fixes the issue. So we get it fixed for free.

0 files changed, 0 insertions, 0 deletions