Simplify stride interface with F or C layout

[ElliottKasoar]

This is not a high priority at the moment, but as discussed in #24, it would be good to improve the user friendliness of the stride functionality, through a F or C layout style and shape/size inference.

This should also be documented in (and potentially help clarify) the discussion about when to transpose arrays, in both the ResNet example and (soon) the long-form documentation.

[jatkinson1000]

If you look at the ftorch.f90 source there is a function t_t_from_array() (which should probably become torch_tensor_from_array) that assumes 'F' indexing and removes the need for the user to pass in the layout argument (used to set strides).

The idea of this is that is creates a tensor that can be read in with the same indexing in Torch without the user having to worry about it.

1. This should be the default option to users and we should be VERY EXPLICIT in the docs that using this function you assume the same indexing in Fortran and Torch.
2. We should look at if we can simplify the examples, making the use of torch_tensor_from_blob with layout an 'advanced' option
3. This should be clearly set out in the README and API docs (see Add Long-form API documentation #53)