Use a single ResizeObserver instance

[ZeeCoder]

Seems like it's more performant and the recommended pattern:
WICG/resize-observer#59

[shiba-codes]

Related – #26

[ZeeCoder]

I'm not convinced yet whether this is an optimisation that's actually worth it or not.
I've only seen a single benchmark that seem to indicate that it might be better to use a single RO for all hook instances,
but it wasn't convincing enough to warrant the associated work.

I'd need a way to test this somehow for myself for a large number of observed elements and measure CPU / memory usage maybe, to see if there's a difference between using one or many RO instances.

[ZeeCoder]

Or if there was a comprehensive test / article / benchmark that demonstrated that using a single RO instance is significantly better, then I would consider it.

[ZeeCoder]

I forgot about this one:
que-etc/resize-observer-polyfill#13 (comment)

Seems like there's something in there, but I'd still need a way to benchmark CPU + memory usage, so that I could compare a results of the current implementation to one that uses a single instance.

[ZeeCoder]

I'd really appreciate someone well-versed in benchmarking UI to help out.