[tabular] Add logging of inference throughput of best model at end of fit

[Innixma]

From user: "It wasn't really clear that predict was going to be going for a long time"

I think we can make this a bit better by mentioning at the end of training the estimated inference throughput of the selected best model, which the user can refer to when gauging how long it will take to do inference on X rows. We have the number already calculated, we just haven't put it as part of the user-visible logging yet.