[Performance] Improvements for QuickView

[redmunds]

These suggested improvement were split off of #4885.

1. Current, line is scanned every time mouse moves more than 1 char. Then, delay 350ms before displaying popover. Seems like we could save some (a lot of?) processing by delaying first, then if mouse is still at same position, scan & display. Note that @peterflynn thinks there may have been a good reason it was done this way, so there may be some gotchas.
2. We check char position to make sure it's in code to popover, but we don't bail out of the loop after we get past current position.
3. Do not scan lines that are over a certain length. CodeMirror does this for code coloring and matching braces due to bad performance of minified files. I think a length of 1024 is a reasonable place to start.

[peterflynn]

This morning, I think we came to the conclusion that the reason for (1) was that it used to start preloading image previews in parallel with the hover delay. However, at some point the code changed and it no longer does that. So there may no longer be any compelling reason to do work on every mousemove.

Re (3), whether line length is a factor is really dependent on the provider. The imagePreviewProvider() only looks at the token the mouse is over, so it will work fine at any line length. So I think we can just put a short-circuit specifically in colorAndGradientPreviewProvider().

[redmunds]

Nominating for Sprint 31.

[jasonsanjose]

@redmunds, now that #5040 is merged, do you want to keep this open to address item #3 in the description?

[redmunds]

Implementing item 3 will cause a reduction in functionality, so that should only be implemented if items 1 and 2 are not sufficient for all use cases. I think we can close this.

[redmunds]

Closing.