Allowing `FunctionOperator`s with unspecified/unused parameters

[gaurav-arya]

Imagine we have a function g(op) -> FunctionOperator(...), which makes a function operator that uses op. If op has u/p/t dependence, we could accept these parameters in g and thread them through to FunctionOperator.

But it can be more ergonomic for op to directly encode the dependence on u/p/t and for the implementation of g to not even have to think about the existence of p and t. The following code would help achieve this:

function update_coefficients!(L::FunctionOperator, u, p, t)
    ops = getops(L)
    for op in ops
        update_coefficients!(op, u, p, t)
    end

    # Allow p/t = nothing to indicate that the parameters are not used by the operator, and hence need not be stored. 
    (L.p !== nothing) && L.p = p
    (L.t !== nothing) && L.t = t

    nothing
end

This came up in the context of JacVec from SparseDiffTools.jl. After discussion with @vpuri3, we've decided to hold off on this change since we can just thread the parameters through for now, but making an issue for the future.

[gaurav-arya]

Edit: L.op already gets updated since it is part of getops(L), so only the unused param task is left -- updated issue title accordingly. Again, we're going to hold off on this for now