fix: more event handling tweaks

[dummdidumm]

• ensure we only have a single document listener per event+runtime
• add <svelte:body> listeners to before_init similar to the document/window elements
• move some code into events.js where it belongs

follow-up to #12376 which isn't released yet, therefore no changeset

Before submitting the PR, please make sure you do the following

• It's really useful if your PR references an issue where it is discussed ahead of time. In many cases, features are absent for a reason. For large changes, please create an RFC: https://github.com/sveltejs/rfcs
• Prefix your PR title with feat:, fix:, chore:, or docs:.
• This message body should clearly illustrate what problems it solves.
• Ideally, include a test that fails without this PR but passes with it.

Tests and linting

• Run the tests with pnpm test and lint the project with pnpm lint

[changeset-bot]

🦋 Changeset detected

Latest commit: bf45ba2

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
svelte	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[trueadm]

I'm unsure about this. Ideally we'd increment a counter for each root, and remove the document listeners when the count goes to zero – otherwise we're retaining a document event listener for no reason.

[dummdidumm]

There's no way that we can intercept unmounting, because it's not required to unmount if everything that has references to the instance is garbage collected.so the trade-off is lingering document listeners vs duplicate document listeners if you mount more than once.

[trueadm]

@dummdidumm What do you mean by intercepting unmounting? If someone doesn't unmount their root, it will cause a memory leak. Tested it just now with this case locally.

[Rich-Harris]

Yeah, people need to unmount their components properly, you can't rely on GC. Though whether or not they bother to do that, given n delegated events and m roots, it's better to have n listeners than n * m.

I added the counter, and if people do unmount their components listeners are once again removed, but if they don't then there's now a limit to the damage they can do, so this seems to be the best of both worlds