the existing numeric tokenizer involved allocations and string parsing. since
the number formats in pdf files are fairly predictable we can improve this
substantially
the file provided in issue #926 contains the following syntax
in pdf object streams:
```
% 750 0 obj
<< >>
```
currently we read the comment token and skip the rest
however this producer is writing nonsense to the stream.
comment tokens are only valid outside streams in pdf files
so we align to the behavior of pdfbox here by skipping the
entire line containing a comment inside a stream which fixes
parsing this file.
* Avoid encoding ASCII in more cases
* Make Space a const
* Use WriteWhiteSpace extension to eliminate possible virtual call
* Use ASCII when encoding constrained character subset
* Simplify pragmas
* Revert Whitespace rename
* Fix using statement order
* Remove obsolete serialization support on .NET
* Remove obsolete serialization support on .NET (part 2)
* Make AdobeFontMetricsLigature a struct
* Make AdobeFontMetricsCharacterSize a struct
* Eliminate allocation in CompactFontFormatData
* Pass TransformationMatrix by reference
* Seal Encoding classes
* Make SubTableHeaderEntry a readonly struct
* Introduce StringSplitter and eliminate various allocations in GlyphListFactory
* Eliminate a few substring allocations
* Use char overload on StringBuilder
* Eliminate virtual calls on stringIndex
* Optimize ReadHelper ReadLong and ReadInt methods
* Add additional readonly annotations to PdfRectangle
* Optimize NameTokenizer
* Eliminate allocation in TrueTypeGlyphTableSubsetter
* Use empty arrays
* Eliminate allocations in OperationWriteHelper.WriteHex
* Use simplified DecryptCbc method on .NET 6+
* Fix windows-1252 encoding not working on net6.0 and 8.0
* Update int buffers to exact unsigned max length and eliminate additional byte allocation
* Fix typo
* Remove unused constant
if we're parsing a known dictionary (e.g. all keys are required
and there are no additional optional keys) and we encounter
an error we provide the possibility to recover by assuming
a dictionary end token after all required tokens are consumed
if parsing by looking for dictionary end failed due to a format
exception
