Improve computation of residuals

[devmotion]

This PR changes the computation of residuals to a slight modification of the implementation in OrdinaryDiffEq (see SciML/OrdinaryDiffEq.jl#574). Moreover, it adds adaptivity parameter delta (see http://docs.juliadiffeq.org/latest/solvers/sde_solve.html#Special-Keyword-Arguments-1) to all adaptive algorithms.

This PR fixes #123, the example

using StochasticDiffEq

f(u,p,t) = 0.4.*u
g(u,p,t) = 0.0.*u
u0 = [1., 1.]
prob = SDEProblem(f,g,u0,(0.0,1.0))
solve(prob, SKenCarp())

works now.

Additionally the PR fixes similar errors for algorithms SRA, SRA1, and SRA2 that were mentioned in SciML/DifferentialEquations.jl#401 (comment), the example

using StochasticDiffEq, DiffEqNoiseProcess

function numerical_sol(solver)
    f(u,p,t) = u
    g(u,p,t) = u
    u₀=[1.0im 2.0im; 3.0im 4.0im]
    W = WienerProcess(0., 0., 0.)
    tspan = (0.0, 1.0)
    prob = SDEProblem(f, g, u₀, tspan, noise=W)
    solve(prob, solver)
end

doesn't fail anymore for these algorithms.