Using SR.jl types in conjunction with DE.jl

[tomkimpson]

using DifferentialEquations
using StochasticRounding
f = (u,p,t) -> (p*u)
prob_ode_linear = ODEProblem(f,Float32sr(1.0)/Float32sr(2.0),(Float32sr(0.0),Float32sr(1.0)),Float32sr(1.01));
sol =solve(prob_ode_linear,Tsit5())

Throws a MethodError: no method matching Float32sr(::BigFloat)

Discussed on Slack - may be able to do something similar to how BigFloats are handled in https://github.com/milankl/SoftPosit.jl

[milankl]

I believe this issue is similar milankl/SoftPosit.jl#65, the BigFloat conversion is no problem to add, with the other methods I want to be a bit more careful.