Fix sliding animation on table elements in Safari

[chrisirhc]

This is a proof-of-concept PR for comments. It is not yet cleaned up. See below for what's incomplete.
This PR fixes #4712 and fixes #4948 .

The approach is as follows:

• Add the ability to add non-animatable CSS (called staticCss in this PR) that is present throughout animation
• Before animation starts, the staticCss is applied through a className on the element
• After the animation is complete, the animation is cleaned up.

For #4712 : Since the overflow property's Animation type is discrete, the behavior is actually defined as changing overflow at the 50% mark if it wasn't overflow: hidden to begin with. However, the sliding animations require that overflow: hidden throughout the animation. This PR moves the overflow: hidden into a staticCss which is applied throughout the animation.

For #4948 : The display property is not animatable, so it can't be applied through animation keyframes. This PR adds a display: block if it detects that the element's display property contains `table. It could instead use a whitelist but I wanted to keep this simple and address #4948 before attempting to address all scenarios that the display property.

I didn't think this was a large enough change to warrant an rfc. Please let me know if it is.

PR is now ready for review.

I plan to add tests since this behavior adds an additional step to animations and needs to be tested.

• Get a rough OK to see if this approach even makes sense before moving ahead with clean up.
• Include a test that fails without this PR but passes with it.
• Fix behavior for out-transitions (this PR does not address staticCss API for that).
• Refactor/clean up style_manager to be able to handle "static" rules cleanly. Currently, this is a hack with some copy-pasta.

Tests

• Run the tests with npm test and lint the project with npm run lint

[flymedllva]

When will the release be released?

[chrisirhc]

Added a test and cleaned up the PR.
Wanted to use window.getComputedStyle to get the computed styles for component, but the tests would fail only in the hydration flow. Not sure if that's because the stylesheet doesn't exist through hydration flow.

[vercel]

@benmccann is attempting to deploy a commit to the Svelte Team on Vercel.

A member of the Team first needs to authorize it.

[rkofman]

Just ran into #4948 and found this old open PR. Would it be useful for somebody to dust it off and get it ready to merge? If not, it would be lovely if it was closed w/ a rejection reason.

[benmccann]

Yeah, as far as I'm aware, the main reasons it hasn't been reviewed is because there are merge conflicts and it's buried down the queue. Sending a cleaned up version would address both of those issues

[Rich-Harris]

Thank you for the PR — closing in favour of #14930 and #14936