Add arithmetic operators for float and double grayscale pixels

[sdebionne]

Is your feature request related to a problem? Please describe.

I was surprised to see that arithmetic operations on pixel compiles but without explicit support from the library. pixels are actually implicitly converted to their underlying type.

Describe the solution you'd like

Provide a set of arithmetic operators, at least for pixel_gray32f_t and pixel_gray64f_t.

C++ Example

#include <type_traits>
#include <boost/gil/pixel.hpp>
using pixel_t = boost::gil::pixel<float, boost::gil::layout<boost::mp11::mp_list<boost::gil::gray_color_t> > >;
int main()
{
    pixel_t a(1.0f), b(2.0f);
    auto c = a + b;
    static_assert(std::is_same_v<std::decay_t<decltype(c)>, float>);   //OK , but is implicit conversion right?
    static_assert(std::is_same_v<std::decay_t<decltype(c)>, pixel_t>); //Fails, but should be OK?
    return 0;
}

Additional context

Discussed first on Slack