I renamed `parsecmaps.py` to `pdftour.py`.  Now it's possible to use
it to navigate the structure of at least one PDF file well enough to
parse a CMap out of it.

This involved adding some stream support, which involved tweaking the
parsing engine a bit.  In keeping with the rest of the
fast-and-loose-exploration nature of the program, it doesn't even
check for `endstream` after the end of the stream, much less `endobj`.

With that, and a bit of tweaking, my `cmaps_for_pages` code from last
week runs now!

Now I can open a PDF file and parse some objects out of it.  Soon I'll
be able to traverse the object graph.

Now we can see what grammar `csranges_to_grammar` has constructed for
us from the CMap file and whether it makes sense.