Remove 'order' arguments from CG operators

[sdrave]

This PR simplifies the implementation in operators.cg by removing the optional integration order arguments from all Operators. All data functions are now evaluated at the entity centers, corresponding to the mid-point integration rule. For the implemented first-order schemes this should be sufficient (assuming regularity of the coefficient functions).

For discontinuous diffusion tensors, it is now in particular clear, that these are approximated by entity-wise constant functions (given by evaluation at the entity centers).

[sdrave]

Any remarks, @renefritze, @ftalbrecht, @deneick?

[renefritze]

Not sure that I follow the reasoning. Isn't this removing high order evaluation of smooth data tensors, which can greatly impact accuracy, just so that the user might not incorrectly evaluate disc. tensors?

[sdrave]

@renefritze, the schemes are all first order. A quadrature is of sufficient order, when the error due to inexact quadrature is of the same order as the error due to discretization. If I haven't overlooked something, the used quadratures should be of sufficient order.

[renefritze]

Ah, sorry, I misremembered then. I thought the CG scheme was implemented with variable order too.