Handle nothing grads for Pairs.data

[ToucheSir]

MWE:

no_kwarg_grad(x; kwargs...) = x[kwargs[:i]]
gradient(x -> no_kwarg_grad(x; i=1), [1]) # MethodError: no method matching getindex(::Nothing, ::Int64) on master

@ChrisRackauckas it also appears that whatever SciMLSensitivity downstream tests are running are inadequate to catch stuff like this. In fact, I'm not sure if any are running given that file still reads DiffEqFlux. What should we replace that line with?

[ChrisRackauckas]

It's just in a different test group. #1296

[ChrisRackauckas]

Bump?

[ToucheSir]

JuliaRegistries/General#67317