Previously, the following piece of code would compile and only
only predictable fail at runtime with assertions enabled.

  Dune::FieldMatrix<double, 2, 3> A;
  // populate A

  Dune::FieldMatrix<double, 3, 2> B_bad;
  // populate B_bad
  A.rightmultiply(B_bad);

It now fails to compile.

Previously, the following piece of code would compile but then enter a
loop, typically leading to a segmentation fault

  Dune::FieldMatrix<double, 2, 3> A;
  // populate A

  Dune::DynamicMatrix<double> B_good(3, 3);
  // populate B_good
  A.rightmultiply(B_good);

This would also happen for the following illegal extension of the above code

  Dune::FieldMatrix<double, 3, 2> B_bad;
  // populate B_bad
  A.rightmultiply(B_bad);

The former now works as expected while the latter fails an assertion.

added void_t (incl. voider-hack for gcc), and template aliases real_t and field_t for convenient access of T::real_type, T::field_type

added void_t (incl. voider-hack for gcc), and template aliases real_t and field_t for convenient access of T::real_type, T::field_type

This commit renames the `vec_access()` and `vec_size()` methods to match
the ones in the `DenseVector` facade class, that is `operator[]()` and
`size()`.

Closes #3

The standard leaves (the effect of instantiating) complex<T> for more
general T unspecified.

While we're at it, kill duplicate tests and add long double to the list

In the current state the scaling parameter for matrix-vector products
(e.g. the alpha parameter for DenseMatrix<...>::usmv) is the
field_type of the matrix. I you consider mixed precision, like in the
xblas library, you would actually want the scaling to be of the same
accuracy as the output vector; an other reason is vectorization for
one matrix and multiple vectors.

The patch contains two changes:

a) the FieldTraits are extended, so that the contain specializations
   for C and C++ vectors
b) the thre matrix implementations are changed, such that they deduce
   the type as FieldTraits<Y>::field_type, where Y is the type of the
   result vector.

In the case of identity matrix the old implementation was laso completely wrong, as it used
Y::axpy to implement usmv and Y::axpy expects alpha to be Y::value_type, so that we already
had a type clash, which just went unnoticed, as we never used mixed-types.