improve use of using inline ...

[erikerlandson]

@armanbilge made interesting use of using inline along with other inlined code, for example here:
https://github.com/gemini-hlsw/lucuma-refined/blob/fa0b3d9a990e1ae94f56f1423473d95e6fcdc5f5/refined/src/main/scala/main/lucuma/refined.scala#L31

This is a discussion thread for re-visiting my thoughts about what (ideally) coulomb expressions "should" compile down to, and how close it is possible to get to that ideal.

I currently create in-line conversion coefficients, so that coefficient_double[U1, U2] compiles directly to a double literal, and in a world where I could get all the definitions maximally in-lined, and the jvm bytecode supported it, a conversion from U1 to U2 would always compile directly to c*v where c is the coefficient constant and v is the original value.

[erikerlandson]

currently I try to use lambda functions as the superclass whereever it makes sense, for example UnitConversion[V, UF, UT] extends (V => V), however Function arguments are not inline, so maybe I'd have to abandon extends (V => V) to do this