(note: cleaved off of #2816, used to be bundled)

Current implementation JsonNodeDeserialization is expensive for deeply nested Object and Array values as it uses recursion: so for each small additional nesting level -- for arrays, 2 bytes to encode [ and ] -- a new stack frame gets created.
In practical terms this means that it is possible to exhaust JVM heap usage with document that has nesting in order of ten thousand(s) levels, depending on settings.

It should be possible to replace basic recursion, however, with iteration, to at least significantly reduce amplification: to prevent cheapest potential DoS concerns.