Update the algorithm for lastPaintedElement

[jasonwilliams]

Related to: #53

This changs the semantic of lastPaintedElement to be the largest element of that frame which was painted, rather than an arbirtary last element which isn't deterministic and can be left open to interpretation.

---

Preview | Diff

[mmocny]

This looks like an improvement in spec wording to me, however, you might be interested in: w3c/largest-contentful-paint#171

Basically, this splits the "LCP selection" from new LCP emission parts of the LCP spec.

You pass a set of new contentful elements (just as paint timing gives) and you get back the single largest one from that set. If you pass an existing LCP candidate to that list, you will only get back a new candidate that is larger than the previous.

For container timing, I think you can:

• Pass all new element paints, without passing a previous LCP candidate, to get the largest for this new paint, with all the same selection and "effective size" semantics (not just rect based).
• For the proposed nested LCP, you can pass the currently running largest LCP candidate so you only get a new one if there is a new largest overall.

[mmocny]

@shaseley

[mmocny]

And you can read how we integrated into soft-navs / icp spec here: https://wicg.github.io/soft-navigations/#report-interaction-contentful-paints-and-soft-navigations