Improve error handling

[natemoo-re]

Changes

• We had a custom page for 500 errors (which was really any error that happens on the server)
• This PR replaces that with Vite's built-in error overlay
• Improves display for generic errors by deriving a code frame automatically
• Updates JSX renderers to "tag" local exports with the renderer matched by our import scanner, which means we can throw much better errors when there's a problem with a JSX file.
  Before
  Errors were uncaught, component ran through many renderers, streaming response would hang indefinitely

After
Errors are caught, component only runs through the correct renderer (which errors helpfully), streaming response closes and Vite's error overlay is shown

Testing

Updated tests for error conditions. Unfortunately was unable to automate error overlay tests due to CI limitations.

Docs

QoL improvement

[changeset-bot]

⚠️ No Changeset found

Latest commit: 1a2a883

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

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

[natemoo-re]

E2E tests are failing because of the issue fixed by #3915!

[bholmesdev]

Some comments, but verrory impressive!

[bholmesdev]

LGTM once smoke passes ✅

[FredKSchott]

Left some comments but not blocking / LGTM otherwise

[matthewp]

I'm not understanding how __astro_tag_component__ is relevant to error messages. Is it? If its for the sake of speeding up rendering I think that should be done in a separate issue.

---

We used to cache the result of .check() in a weakmap. This worked with any framework and didn't rely on AST parsing. I can't remember why we removed that, but if this is being done for performance I'd rather rethink about that approach.