toJSON can cause stack overflow

[AndreasMadsen]

Because toJSON works recursively dprof can reach a stack overflow. This could be fixed by flattening the format, such the root is a uid map and node.children is a uid list.

Root {
  version: String, // the version of dprof there generated this JSON file
  total: Number, // execution time in nanoseconds
  root: {
    Number(uid): Node, ...
  }
}

Node {
  name: String,      // Handle name of the async operation
  stack: [           // Contains the stack leading up to the async operation
    {
      description: String,
      filename: String,
      column: Number,
      line: Number
    }, ...
  ],
  init: Number,      // Timestamp for when the async operation is requested.
  before: [Number],  // Timestamp for when the callback is about to be called.
                     // This is an array because a callback may be called more
                     // than once.
  after: [Number],   // Timestamp for when the callback is finished.
                     // All timestamps are relative to the process startup time
                     // and the unit is nanoseconds.

  unrefed: Boolean,  // `true` if the handle will not keep the event loop open.

  children: [        // Shows async operations created in the callback
    uid, ...
  ]
}

[AndreasMadsen]

It may be better to use a 2d array, such it easily can be transferred to and from a Map. (from: Array.from(items.entries()), to new Map(Root.root)).

Root {
  version: String, // the version of dprof there generated this JSON file
  total: Number, // execution time in nanoseconds
  root: [
    [Number(uid), Node],
    ...
  ]
}