Enhance sage.tensor to support general rectangular arrays

[amanmoon]

Problem Description

The current implementation of sage.tensor only supports tensors of the form (M^*) ^k (x) (M) ^l
We propose to lift this restriction and enable the creation of more general rectangular tensors.

Proposed Solution

1. Make the self._fmodule argument in the TensorFreeModule.__init__ and FreeModuleTensor.__init__ class to store all the modules used in the creation of the tensor product module as a tuple.
2. Modify all the methods in sage.tensor keeping the above changes in mind.
3. Make the self._tensor_type argument conditional, it will not be initiated when the user creates tensor product of different modules.

The final API should look as follows:

N = FiniteRankFreeModule(ZZ, 3, name='N')
M = FiniteRankFreeModule(ZZ, 4, name='M')
T = tensor([N, N.dual(), M, M.dual()]) # this will be of type TensorFreeModule_with_category
t = T.element() # this will construct an element on T