feat: Add iterator methods to Query

[segevfiner]

Add iterator methods to Query

Still need to do capturesIter and CapturesIterator, typings, etc.

I'm not that happy about the code duplication though... We could probably implement matches and captures in terms of matchesIter and capturesIter to reduce it. But that might be slower due to all the additional calls that would entail between C++ and JS, and transferring the nodes one by one.

P.S. LookaheadIterator[@@iterator] can probably be implemented in C++, and ts_query_cursor can probably be allocated on Query instead of in AddonData if we want too.

Fixes #178

[amaanq]

I think it'd make more sense to just implement matches and captures with IterableIterator by default, what do you think?

[segevfiner]

I'm worried about backwards compat abd performance. Which is why I kept the existing ones for now.

[amaanq]

If performance is a concern, why even add this then? What's the benefit/purpose exactly? (really curious because I'm all for iterators)

[segevfiner]

Memory consumption. As using iterators means we don't need to hold all results in memory at once. We still need to measure if it's really significantly slower or not. I'm not sure if it's really slower. But if it is, it is always possible to chunk the iteration internally so we don't transfer one by one.

[maxbrunsfeld]

From my experience, using iterators is going to be much less efficient, because there are many JS -> C++ calls, not just one. I don't think we should make this change at all, unless there is some very specific benchmarks demonstrating that it solves a performance problem.

[segevfiner]

Very large result sets are one reason. But we may need batching of results for this to be efficient. And benchmarks obviously.