refactor Complex.frag to support both float and double (#91)

refactor Complex.frag to support both float and double
new DoubleMath.frag for double precision trig etc
- copy/pasted from old Complex.frag
- builtin* captures for overloading (built in functions can be redefined but not overloaded directly)
new ComplexBase.frag for (dual) complex maths using REAL, VEC2, VEC3, VEC4 (number type agnostic)
Complex.frag is now a simple skeleton, that includes the above two. If double is supported (tested by VERSION >= 400) then REAL etc are redefined to the double precision variants and ComplexBase is included again.
User frags that wish to support both float and double (using highest precision available) should set #version and use REAL VEC2 etc. Then the only thing that needs to be changed to support old hw is the #version.
User frags that wish to use both float and double at the same time should use float, vec2 for low precision and REAL, VEC2 for high precision. This will gracefully downgrade to float for both if double is not supported.
If user frags absolutely need double precision, then #version 400 compatibility and just use double, dvec2, etc.
Tested briefly with BurningShip.frag, adding a #version line doesn't break it horribly. DoubleTest.frag also works fine.

3Dickulus / FragM