OpenSystemsDevelopment / qpdf

Move all array-specific methods to new class qpdf::Array.

Move all dictionary specific methods to BaseDictionary. Create new subclass
Dictionary to allow QPDFObjectHandle to access the moved methods.

Replace data member oh with accessors.

Remove redundant overloads from QPDFObjGen::set.

The new class is only usable as base classes and does not support direct
instantiation or upcasting.

Add operators to convert to bool, QPDFObjectHandle and QPDFObjGen.

Derive QPDFObjectHandle from BaseHandle and make final.

#1349 introduced a limit on the maximum size of arrays and dictionaries
contained in objects that generate errors during parsing, and #1354
reduced that limit to 5000 objects. However, the limit was only imposed
once a further error was encountered.

Stop adding objects to containers once the limit is reached.

Fixes oss-fuzz issue 398060137

Remove member encoded and encode strings only when required.

Encoded is only used when writing JSON. Since most JSON writing is now
done on the fly without creating a JSON representation, creating and
storing the encoded string in almost all cases no longer serves any
purpose.

If duplicate keys are encountered, overwrite earlier values with the latest
value.

Except for virtual functions to avoid the need for users to update code.

Also, move Buffer::Members definition to Buffer.cc

Fix the logic to accept a top-level `zopfli.h` header, as that is
the location used by upstream's build system.

Removing it was ABI-breaking. I must have done it forgetting Pl_DCT.hh
was in the public API.

This improves indentation of long strings. This commit also fixes some
trailing whitespace in ChangeLog.

Reduce the container size for which a single bad token will cause a failure
from 100,000 to 5,000.

Count missing dictionary keys as errors.

If a flate memory limit is set, reject decode parameters that would cause
the limit to be exceeded and treat stream as unfilterable.

Fixes oss-fuzz issue 394463491

The stream filter is only used if the stream is filterable, therefore
there is no point continuing to set decode parameters once it is recognized
that the stream is not filterable.

Fail if a bad token is encountered while parsing an array or dictionary
with more than 100,000 elements.

Fixes oss-fuzz case 388571629.

Fixes oss-fuzz case 394129398.

Issue arose from chaining multiple runlength filters and inflating a
compressed stream of ~100 bytes to several gigabytes.

There is no obvious fix without imposing an arbitrary implementation limit
and therefore potentially excluding valid PDF files.

This requires a special build option.

Check for direct outline items.

Fixes oss-fuzz case 	389339260.

Fixes oss-fuzz cases 376305073, 389974979 and 391974927.

When recovering XRef streams, start with the stream with the largest
/Size rather than the largest offset.

Also, if reconstruction fails to find a trailer with a valid /Root entry
search for a root object.

Change QPDFWriter stream_decode_level default to qpdf_dl_generalized (fixes #1286)

runtime error

Also, fix disabling of preserve_encryption to be ignore
stream_decode_level, but disable preserve_encryption if compress_streams is
false.

Fixes #1286