Generate CP ALS Runtime Comparisons Across Tensor Sizes and Dimensions

[ntjohnson1]

Right now small tensors (in total entries and number of dimensions) appear to be faster in pyttb. However, for larger ones rusty_tensor (via PyO3 bindings) seems to win. We should figure out a range of sizes to compare against to look at the scaling trend. We should probably also investigate instrumenting rusty_tensor to see where we are spending the most time. I know we are doing a few unnecessary copies (at the rust/python boundary and at struct creation), but that might be a trivial amount of time for the scale of problems for now.

A couple unscientific comparisons on my machine:

Pyttb time: 3.5205395221710205
Rust time: 2.277113914489746
Shape: (100, 100, 100, 10)    

Pyttb time: 30.862141132354736
Rust time: 24.287920236587524
Shape: (100, 100, 100, 100)