Performance improvements for the PDOM

[zepumph]

Over in https://github.com/phetsims/ratio-and-proportion/issues/356, @jessegreenberg and I found a couple of places where we could improve scenery's performance as it pertains to the PDOM. Mostly this has to do with the frequency and quality of HTMLElement manipulation. Here is the list we came up to try out and see if performance improves. I was also thinking that potentially we could make some examples, perhaps with animation (outfit blast/?) to expose some performance-constrained cases.

• Separate out position-focused style from the a11y-pdom-element rules,
• remove calls to setClientBounds style in PDOMPeer on non positioned elements.
• Look into batching attribute setting to a single time per frame. (right now it is often based on model Properties, and may involve multiple, synchronous, DOM operations per frame. I'm not positive this is the case; it will need more investigation.
• Try to minimize the amount that we ask for data from the DOM element, and instead just try to keep that in the PDOM trait.

We did not feel like this work blocked Ratio and Proportion RC, but it should have some time spent on it soon. At least for the low hanging fruit.

[zepumph]

I see about 1000 calls to setting textContent on a dom element when spamming the RAP hand up and down. I believe that is a synchronous step to update attributes. I know we don't want to kick off a total rebuild during next update display, but perhaps we could still only update attributes/generally mutate the peer on update. This will likely involve creating "dirty" flags, but could be reasonably possible/simple