Improve the performance of the Katz Centrality computation.

[Paulo-21]

What is the expected enhancement?

Improve the performance of Katz Centrality.

I saw that for every iteration of all node the beta values was looked from a hashmap index by the node index (usize) exactely like the x vector .
I think the hashmap could be replace with a simple vector, this lead to better performance.
This could be done without breaking the api.