OpenSystemsDevelopment / qpdf

This change updates all occurrences of `std::endl` to `\n` across various files for improved efficiency and consistency. It reduces unnecessary flushing of output buffers associated with `std::endl` while maintaining the same output format.

Simplified checks for empty containers and strings across the codebase using the `.empty()` method. This improves code readability and adheres to best practices for checking emptiness.

Converted multiple occurrences of `count()` to `contains()` throughout the codebase where the goal was to check key existence in containers. This improves code readability and aligns with modern C++ practices, particularly with C++20, making the intent more explicit and potentially aiding performance.

Only top-level XMP metadata is supposed to be left unencrypted. All
other metadata is not treated specially.

- Add missing export for operator QPDFObjGen.

- Fix handling of indirect references. The use of the type_code method
  in various switch statements bypassed the special handling of
  ::ot_reference objects. Replace with new method resolved_type_code.

Add tests for ::ot_reference objects.

Exercise stream containing objects with no white-space between them.

Also add new fuzz test case.

The code accepted values other than /Yes but still used /Yes as the
checked value instead of obeying the normal appearance dictionary.

By combining --linearize with --compress-streams=n, we ensure that no
new compressed data will appear in linearized output, which makes the
output independent of zlib's output. There are other tests to ensure
that linearization works correctly with compression. This commit
involves changing some test outputs and test code as well just
updating test suites.

Also, fix test cases.

...where the first "number" is an indirect object that happens to be a
number.

Exclude uninitialized, destroyed and reserved objects.