Provide some measurements each sample

[PaulBone]

Please excuse me if we do this already.

We could provide some easy to obtain measurements each profiling sample and graph them over a shorter time range than markers normally let us measure things.

I can think of the current amount of allocated memory/number of objects in the JS and other heaps, but there are probably plenty of other ideas. If we think it's useful and it's not too invasive we could query the number of cache hit/smisses/instructions retired or simply how much time was spent "on CPU" vs idle.

As always, I can help with the GC stuff.

┆Issue is synchronized with this Jira Task

[julienw]

Looping in @tarekziade in this discussion as this is/can be in the scope for his current project and more generally what he thinks of this.