Reverse cleanup of #13543

[paulmasson]

@mrdoob the supposedly minor cleanup on #13543 forces the end user to work in a way that IMHO disrupts how a mathematically inclined person thinks about functions. A parametrization function returns the coordinates corresponding to the input parameters: it doesn't modify the parameters. I should think someone using ParametricGeometry would expect a function to behave like a mathematical function.

With the cleanup reversed, the end user still has the option to modify a third argument if desired for efficiency, but doesn't force that person to think non-mathematically. I would really like to know whether or not this cleanup was strictly necessary from a performance point of view.

[WestLangley]

three.js no longer will instantiate a new Vector3 every time a user evaluates a function.

If the user is able, the user can instantiate one Vector3 instance and reuse it.

This pattern is extensively used in this library. For example:

https://github.com/mrdoob/three.js/blob/dev/src/math/Ray.js#L40-L51

[pailhead]

Not all functional paradigms can be applied to three. This is an anti-pattern with threejs

[Mugen87]

Closing. Merging this would be a return to the old system.

[paulmasson]

Then there should be an indication for the end user as to why it simply fails: #14085.