Custom CFL function and body force

[b-fg]

These are small improvements to allow greater control of the time integration loop. Users can define custom CFL functions and pass arbitrary body forces.

A trivial case is to use a constant time step

CFL_ct(a::Flow) = 0.25
sim_step!(sim; CFL_f=CFL_ct)

Also body forces can be passed to the RHS easily with

random_force = randn(size(sim.flow.f))
sim_step!(sim; bf=random_force)

[marinlauber]

Looks good! I was just wondering what happens for the first time step? We force it to be 0.25 when the Flow is constructed

WaterLily.jl/src/Flow.jl

Line 110 in 775c6bc

function Flow(N::NTuple{D}, U; f=Array, Δt=0.25, ν=0., g=nothing,

. Maybe we want this new CFL function to change that as well?

Any idea why the tests are failing?

[b-fg]

The first time step can already be modified by the user, so no worries on that. One can just pass Simulation(...; Δt=CFL_f(0.0)) and create it's own CFL_f(t::Number) function, in addition to CFL_f(a::Flow).

I am not sure why the Julia 1 x86 tests are failing. It seemed to be a CI problem when loading the dependency packages, and not a problem of our tests.