add ndarrays.flags.f_contiguous

[ev-br]

Implement f_contguous flag, also heuristics for owndata and writeable flags.

[ev-br]

this is of course just torch.cumprod(self._tensor.shape, 0); my instinct from numpy usage is to avoid array operations for small tuples etc.

[lezcano]

A simpler implementation in numpy would be x.T.is_contiguous(), using the terrible semantics numpy gave to the T attribute. Either sgtm

[ev-br]

A simpler implementation in numpy would be x.T.is_contiguous(), using the terrible semantics numpy gave to the T attribute.

Exactly. I considered doing just that, but this would need to live in the wrapper level and indeed rely on numpy semantics, and we're trying to push things to the torch level, are we.

[lezcano]

if things are implemented in the wrapper, we can certainly use those to implement others as those will redirect ultimately ti PyTorch ops, so that's completely fine