Add new benchmark for sparse AD on simple loop written diffusion operator

[yolhan83]

This PR adds a new benchmark to test sparse automatic differentiation (AD) methods on a simple diffusion operator implemented using a loop. It evaluates various AD backends, including Enzyme, ForwardDiff, Mooncake, and sparse configurations with TracerSparsityDetector and GreedyColoringAlgorithm. A manual Jacobian is provided for validation, and the benchmark framework compares performance and accuracy for a random input vector of size 1000.

[ChrisRackauckas]

Error: ArgumentError: Package DifferentiationInterfaceTest not found in cur
rent path.
- Run `import Pkg; Pkg.add("DifferentiationInterfaceTest")` to install the 
DifferentiationInterfaceTest package.

The project/manifest need to get updated.

[yolhan83]

It should work now, thank you

[yolhan83]

wait were the tests suppose to be in 1.10.5 ? The other sparse test runs fine on 1.11 though

[ChrisRackauckas]

I'm not sure if the table is nicely displayed? But we'll see in the dev version of the docs. The results look nice and useful though.

[yolhan83]

If it's not there is the parse from the Julia AD Benchmark which can be used, will see