This PR fixes the dynamic programming algorithm. In particular, we ensure the end point of the aligned curve corresponds to the end point of the curve. I'm afraid we are not there yet, but hopefully this is a first step towards correctness.

Moreover, I've tried to improve naming in order to make the code more readable, but I'm afraid it is still too hard to follow.

Known issues (also occur in previous implementation)

The dynamic programming outputs a squared dist. After alignment, we can also compute a distance using the total space metric (this is the distance quotient metric outputs).

For more complex cases (thinking about birds), I've realized these distances are similar, but slightly off. Moreover, when we swap point and base_point, the dynamic programming squared dist is the same, but the distance we compute with the total space is not necessarily symmetric (though close).

For simpler cases as the curves generated by random_point, the results are better (correspondence between DP and total space; almost symmetric distance).