fix(troika-three-text): remove IIFEs for tree-shaking

[CodyJasonBennett]

Fixes #215. IIFEs cannot be statically analyzed for tree-shaking -- their scope is forced into the bundle regardless of usage.

[lojjic]

🤔 IIRC, those IIFEs were actually necessary at one point for proper tree shaking; there was something in them (prototype mutation, perhaps?) which was confusing bundlers at the time and the IIFE with PURE comment worked around it. Maybe that's not a problem anymore, or it's just one specific bundler with the issue? (I think I was using Rollup but my memory's hazy.)

Assuming you've verified this fixes treeshaking with the drei stack, I'd be ok accepting the change. Worst case we find out it breaks treeshaking elsewhere and have to come up with another fix.

[CodyJasonBennett]

I can verify this on my end today/tomorrow. This should be released before my other PR so it can be backported. Pure comments are needed for all global side effects, but we've sorted this downstream if three was a factor.

[CodyJasonBennett]

There was indeed a prototype mutation for r109 compat. I've moved it to the class as an overload in b55a0c8. This or similar work-arounds <r125 can be removed following #215.

I've verified tree-shaking works with https://npmjs.com/agadoo. I will manually review with Rollup, ESBuild, Webpack, and Parcel. Are there any tools I'm missing? Happy to create a test repo for future reference or integration testing.

[CodyJasonBennett]

I created a generic testing repo and a branch with a local copy of this PR at https://github.com/CodyJasonBennett/treeshaking/tree/troika. Tree-shaking seems to work everywhere but Parcel which doesn't seem to tree-shake with the given configuration. I tested between Rollup, ESBuild, Vite, Webpack, and Parcel.

[lojjic]

@CodyJasonBennett We've got some merge conflicts now, could you please resolve them? I've merged #225 so the workarounds you mentioned can probably be removed too.

Edit: nevermind about there being conflicts, I was misreading. We can still remove the workarounds though?

[lojjic]

Actually there may be some mis-resolved conflicts -- I'm seeing a return from PlaneGeometry to PlaneBufferGeometry in the diff...?

[CodyJasonBennett]

Resolved them.

[lojjic]

Thank you!

[lojjic]

v0.47.0 is published.