Improve performance by storing LSP traffic on the LSP object itself when converging the model

[tim-fiola]

For large models (# LSPs is large, say 11000), this call below takes a long time to run:

lsp.traffic_on_lsp(model)

I think we can get around this by storing info when we converge the model. I need to look at the source code more, but perhaps adding a _traffic_on_lsp list attribute to the RSVP_LSP when converging the model to store the demands and proportions on the RSVP_LSP so it does not have to be recomputed again.

I'm not sure this is possible, but look into it.