OpenSystemsDevelopment / qpdf

While scanning the file looking for objects, limit the length of
tokens we allow. This prevents us from getting caught up in reading a
file character by character while digging through large streams.

A stray semicolon caused a condition to be incorrectly applied during
stream length recovery.

Files written in PCLm mode have to be created in a very specific way.
qpdf doesn't know how to create PCLm files from scratch. All it knows
how to do is to write an already valid file in a suitable way.
Therefore there is no command-line support for PCLm.

There is no need for a --precheck-streams option. We can do the
precheck without imposing any penalty, only re-encoding the stream if
it fails the first time.

This commit adds several API methods that enable control over which
types of filters QPDF will attempt to decode. It also adds support for
/RunLengthDecode and /DCTDecode filters for both encoding and
decoding.

There was an unintended recoverable error in a test file. It wasn't
hurting anything, but it was obscuring the actual intent of the test.

This is in preparation for implementing page groups.

Add a newline unconditionally before endstream even if a newline was
already written as part of the stream data.

Bad /W in an xref stream could cause a division by zero error. Now
this is handled as a special case.

Remove problematic test files

Also accept more errors than before.

Eliminate PCRE and find endobj not preceded by endstream. Be more lax
about placement of endstream and endobj.

Sometimes we want to ignore bad tokens rather than having them throw
an exception. A coverage case is commented out here and added in a
later commit.

Also fix a bug resulting from incorrect use of PointerHolder because
of this unused parameter.

Passed arguments to the constructor in the wrong order.

main() had gotten absurdly long. Split it into reasonable chunks. This
refactoring is in preparation for handling splitting output into
single pages.