consider moving subtensor_utility.hpp as member functions

[exclowd]

I was thinking if we could do away with the subtensor_utility.hpp and move all the utility functions to private helper functions, reasoning being:

1. Lack of other such files in the codebase
2. Computing strides and extents may be better to think of that way, because they aren't general enough to something like detail::eval.

What do you think about this, @bassoy?

[bassoy]

Agree detail is a good namespace for that!

Those functions could be private member functions of the current subtensor. The question is what happens if you create a new subtensor type. Then you would need to implement those again which violates DRY. So, I propose to keep subtensor_utility.hpp and only remove it when other subtensor types don't need them.