[Feature Anywhere] Dynamically update vega spec to show VisLayers

[ohltyler]

As a follow-up to #3122 , after VisLayers are collected, we will want to pass them into the relevant render function to be processed. This will primarily involve making changes to the vega spec, such that a VisLayer will require adding a new layer that will be visible on the chart (see mockups in #2088 as an example).

[ohltyler]

There are a few options to add events / VisLayers on the chart:

1. adding a layer to the existing view and placing at the bottom of the chart. X-axis can be offset to provide more room for the event marks
2. adding an entirely new view and connecting both via vconcat

Pros/cons of each:
Option 1 - pros:

• less code changes and fundamental changes to the existing spec
• can utilize the existing autosize field to dynamically fit the vis embeddable container
• no need to maintain multiple x-axes configurations & ensuring they line up correctly

Option 1 - cons:

• inconsistent with existing experience (TSVB)
• potential obstruction if the lines come close to the x-axis

Option 2 - pros:

• dedicated space (including empty state) for events
• guaranteed that there will be no obstruction between values & event markers

Option 2 - cons:

• unable to use autosize field due to limitations (note this is what is currently used in the VegaParser)
• difficulty / potential edge cases in maintaining x-axis consistency across charts
• complex logic for trying to generate the size on-the-fly based on the number of views in vconcat/hconcat

[ohltyler]

Resolved via #3145