Improve perf of f-order array operations

[bluss]

C-order is the default order. The following functions need adaption for f-order:

• [x] map
• [x] to_owned
• [x] zip_mut_with broadcast from 0
• [ ] zip_mut_with general
• [x] scalar_sum
• [x] ~allclose~
• [ ] eq

allclose, to_owned, map, scalar_sum can all start with implementing an agnostic version of as_slice()