Changing convergence criterion to match CRAN package?

[rybern]

Hi Brett,

What do you think about changing the convergence criterion to match the CRAN package? It's described briefly here on page 2 (see the "thr" parameter).

Is your alpha parameter is the same as their rho parameter?

Thanks for looking at my other pull requests. I might make a couple more, I hope I'm not annoying!

Edit: To clarify, I'm happy to make the changes, I'd just like your opinion on the idea

Thanks,
Ryan

[bnaul]

Sure, makes sense to me; I think that's just a scale-invariant version of the same criterion, which seems like a desirable property.

Thanks for the contributions! Keep 'em coming 👍