Use FastBroadcast for multithreaded broadcasting?

[glwagner]

For some time we've lamented that while FFTW is multithreaded, broadcasting operations such as

FourierFlows.jl/src/timesteppers.jl

Line 109 in e059ba9

@. sol += clock.dt * (equation.L * sol + ts.N)

are not.

FastBroadcast.jl provides an answer:

https://github.com/YingboMa/FastBroadcast.jl

It seems that even without multithreading, it can speed things up.