also adds a gitignore entry for the 'benchmark' subfolder in tools where custom benchmarking applications can be built and run without being added to source control.
previously the only way to test if a password was correct was to supply a single password and throw if the value was incorrect. this was slow. now parsing options supports a list of passwords as well as a single password option (which is equivalent to a list with a single item). these passwords are all tested at the same time and an exception is only thrown once all passwords are tested.
* while the pdf specification says stream data should follow a newline following a stream operator some files have only a carriage return following the stream operator.
* since comment tokens may appear inside an array or dictionary we ignore them if they occur here since they will break interpretation of the dictionary or array contents.
previously we just treated character codes as glyph ids when getting the bounding box from the truetype font program itself. this change uses the code for character code to glyph id mapping from pdfbox, with some changes, to retrieve the correct bounding box where possible. since this relies in some places on using the unicode value or name, rather than character code, we add a cache to the individual truetype fonts to store the character code to unicode mapping which should have the benefit of improving performance.
* cid cff fonts have multiple sub-fonts and multiple private dictionaries, in addition to a top level font and private dictionary. this fix uses the specific sub-dictionary when getting local subroutines on a per-glyph basis.
* chinese, japanese or korean fonts can use a ucs-2 encoding cmap for retrieving unicode values.
* add support for the additional glyph list for unicode values in true type fonts. adds nonmarkingreturn mapping to carriage return.
* makes font parsing classes static where there's no reason for them to be per-instance.
