feat(get-element-stack): performant api to replace document.elementsFromPoint

[straker]

Just making a pr with the new api for now since it's quite big in terms of code size. Once this is merged I'll make another pr to use it.

The code essentially does the following:

1. It creates a 2d grid of the page and adds every element to every cell of the grid it belongs to. If the element creates a scroll region, then a new grid is created for that element and all its children
2. As it adds elements to the grid, it calculates the stack order of the node using an array of numbers. Each entry in the array is a new stacking context and the number represents the z-index of the element at that level.
3. When the stack is requested for a node, it starts at the grid the node belongs to and calculates the row/col of the center point of the element. It then gets all nodes that are part of that row/col and performs a collision detection against the element and the point. If the grid is a sub-grid, it then goes to the parent grid and gets the stack at that level
4. Finally it sorts all the returned nodes based on their visual order

I also added cache functions to virtual node as we do these lookups a ton throughout the codebase.

In terms of performance. Running this new function on every node on https://developer.mozilla.org/en-US/docs/Web/HTML/Element/template compared to document.elementsFromPoint and our color.getRectStack function (which is what uses document.elmentsFromPoint under the hood for color-contrast checks), the average of 5 runs are:

• color.getRectStack: 143.47900002030656ms
• dom.getElementStack: 53.457000001799314ms

Running it on https://web.archive.org/web/20190613132353/https://giveawaylisting.com/ (have to use web archive result now that the site is going down) results in:

• color.getRectStack: 143147.4000000162ms
• dom.getElementStack: 2190.5150000122376ms

NOTE: Jason's suggestion saved ~20% time (500ms on giveawaylistings.com)

Reviewer checks

Required fields, to be filled out by PR reviewer(s)

• Follows the commit message policy, appropriate for next version
• Code is reviewed for security

[stephenmathieson]

Nitpicky stuff. Will review in depth at a later date.

[scurker]

Some minor nitpicks and micro performance suggestions. Overall, good work here!

Resolved

[WilcoFiers]

Half a review

[dylanb]

Need to add support for the other type of replaced objects - see comment

[straker]

We decided to hold off on supporting CSS rotation for now and add it as a separate PR / function.

Resolved

Resolved