Missing passed on function pieces for ODEForwardSensitivityFunction

[oxinabox]

when you construct an ODEForwardSensitivityFunction from a ODEFunction not all all argument are preserved.
Preserving them is largely a matter of working out when the versions of them for the extended system of equations is, given that one already has them for the initial system of equations.

They are passed onward in this line of code:

SciMLSensitivity.jl/src/forward_sensitivity.jl

Lines 411 to 417 in 6cd2744

	sense = ODEForwardSensitivityFunction(f, f.analytic, nothing, f.jac, nothing,
	nothing, nothing, f.paramjac,
	nothing, nothing,
	uf, pf, u0, jac_config,
	paramjac_config, alg,
	p, similar(u0), mm,
	isautojacvec, isautojacmat, f.colorvec, nus)

Each nothing is something that is missing but not all of them actually matter.

I am listing the ones I have run into that actually cause some (mostly minor) issues here, so we can think about fixing them.

• tgrad
• jac_prototype

These are probably good small tasks for people new to the package who are familiar with the math.
A nice point of not totally trivial but not so hard. And not really subject to much design debate.

[Spinachboul]

Can't we just replace the nothing arguments with the ones that actually matter.
Personally I feel these would be important:

1. sparsity: We can use sparsity pattern of the jacobian matrix which can lead to efficient calculations
2. original_jac: This can allow the ODE solver to track the changes in the original function during integration process, also helpful for adaptive step size control!

@ChrisRackauckas and @oxinabox