Avoid recursion when parsing documents

[walruscow]

Deeply nested documents can cause a stack overflow due to recursion. Avoid that by using a heap based stack instead.

This includes a first commit to infer the sibling index from the parent, so that the iterative approach is simpler

• Tested with a 1mb document with 100k nested nodes, now no longer crashes
• cargo test

This would resolve issues #66 and (arguably) #68

Since there is no maximum depth option added here, it is possible to go fully OOM, but much larger documents are supported than previously.