Add covariant derivative/connection support.

[rj-calvin]

For a given vector , we need a method for calculating the covariant derivative given by:

The covariant derivative should be implemented as an operator, however, the contraction of the input vector with the Christoffel symbols implies that the covariant derivative is not compatible with the current API for tensors.

I propose to simply model the covariant derivative as a tensor operator similar to the DiffOperator in partial.py where it is a subclass of ~sympy.tensor.tensor.TensExpr and defines the __mul__ and __rmul__ magic methods to return an expression of the form provided by the equation above.