Feature grad update

Summary

Avoiding re-allocating cached gradients

Resolved bugs (If applicable)

Enabling recalculation for moving mesh cases.

Details of new models (If applicable)

No new inputs. Re-uses the storage where applicable, including mesh motion.

Risks

• Only used if gradient caching enabled in fvSchemes.
• Tested to give bit-identical results for

Gauss
cellLimited
cellLimited<Venkatakrishnan>
edgeCellsLeastSquares
fusedGauss
fusedLeastSquares
leastSquares
pointCellsLeastSquares

and their derivatives (e.g. edgeCellsLeastSqaures). Currently not natively done is iterativeGauss - ideally would also require caching of face based gradient.

• caching can easily be enabled for all variables in system fvSolution

cache
{
    // Cache all grads
    "grad(.*)";
}

• Tested for memory leak (only a single scheme since caching is generic)