[extrema][optimize] Update max and min functions to allow any order and any axis + some optimization work

[forfudan]

This PR updates the numojo.math.extrema module and performs some other optimization work:

• Update max() and min() to allow both C and F order arrays and by any axis.
• Unify all the overloads and function signatures. The maxT() and minT() are removed.
• Update the max() and min() methods for NDArray type.
• Some other optimization work, including:
  • Use // and % to replace divmod() in all cases.
  • Use a.size attribute to replace a.num_elements() method in all cases.
  • Remove unnecessary copy of memory in the apply_func_over_axis functions.
  • Increase the speed of nditer by not re-constructing the strides in every loop.

[MadAlex1997]

Good work on vectorizing the extrema functions with axis properties. I also like that we are now mirroring numpy with size being a field and the default way to get the total number of elements.

[forfudan]

Good work on vectorizing the extrema functions with axis properties. I also like that we are now mirroring numpy with size being a field and the default way to get the total number of elements.

Thanks! I think with this apply_func_by_axis function, we can quickly implement all core ndarray methods which take axis as an argument. Maybe this can be a milestone for v0.6.

[MadAlex1997]

Good work on vectorizing the extrema functions with axis properties. I also like that we are now mirroring numpy with size being a field and the default way to get the total number of elements.

Thanks! I think with this apply_func_by_axis function, we can quickly implement all core ndarray methods which take axis as an argument. Maybe this can be a milestone for v0.6.

Between that and the 25.1 Mojo update it seems like an appropriate time for main to go to 0.6.