Introduce StackDepthGuard class to check for stack depth in CoreTokenScanner and fix #1217

[BobLd]

I'm not a huge fan of having to introduce a new class in order to track stack depth, but without an in-depth refactoring of CoreTokenScanner, this is needed to fix #1217 This approach could hide stack overflows that could be properly addressed

@rhuijben just to highlight the change since I believe this would need to be taken in account in the refactoring you worked on. The existing implementation of Array and Dictionary Tokenizer classes are the root cause of the issue, as they recursively call CoreTokenScanner.MoveNext(). Finding a way to be able to track stack depth without having to rely on StackDepthGuard would be a better approach.

@EliotJones if you have an opinion on the change.

I want to wait for the next release of PdfPig before release this change

[rhuijben]

Ack.

I would make the Enter() return an IDisposable so you can use using on it without having to use finally yourself.