Reshape and Transpose

Implement two utilities for tinygrad tensors:

• flatten_then_reshape(x, new_shape) - first flatten x to 1-D, then lay the elements out in new_shape. The total number of elements must match.
• transpose_last_two(x) - swap the last two dimensions of x. This is how you turn (batch, in_features, out_features) into (batch, out_features, in_features).

Both functions take and return tinygrad Tensor objects.

Hints

• Tinygrad tensors have a method for collapsing all dims into a 1-D layout, and a method for rearranging into a new shape, as long as the total element count is preserved.
• A different method swaps two axes. It accepts negative indices, so the last two axes can be referenced as -1 and -2 regardless of how many leading batch dims the tensor has.