Replace `vec` methods with operator overloads

[no-lex]

The current class vec, while useful, requires the user to use an unusual means of calculating operations: named methods corresponding to the desired operation. However, C++ gives us operator overloads, and that is the way these types of operations should be conducted. Operator overloads don't require the developer to memorize what the method names are: you just have to call on the well known operators themselves (* + - / += -= > < etc.).

Some things to be aware of when rewriting methods:

From rendersky:

   vec lambda(680e-9f, 550e-9f, 450e-9f),
       betar = vec(lambda).square().square().recip().mul(1.86e-31f / atmodensity),
       betam = vec(lambda).recip().mul(2*M_PI).square().mul(atmohazefade.tocolor().mul(atmohazefadescale)).mul(1.36e-19f * std::max(atmohaze, 1e-3f)),
       betarm = vec(betar).div(1+atmoclarity).add(betam);

We see spurious vec() constructor calls here because vec has no equivalent of the + operator, only the += operator. These vec() calls construct a new object to += on so that the original is not modified. Of course, these should be replaced by + when encountered.

In general, the operator methods already defined are of the += type: they modify the object in place. This includes some operators such as ** which don't even have such operation/assignment equivalents.

The mul(vec o) operator is not the dot operator: mul returns the element-wise multiplication of this with o.