Suggestion: Improve performance through regex memoization

[daniellubovich]

Hey! Recently I've started to hit some performance bottlenecks that seem to be stemming at least partially from route-recognizer. It might be the case that the root of the issue is due to either 1) My application being too large, or 2) ember-source not using the recognizer in the most optimal way, so take whatever I suggest with a grain of salt.

I ran some performance checks when running an acceptance test in Ember. This test loads up all of our routes (we have a lot) at the beginning of the test, and it takes a significant chunk of time. Around 20%.

While I do think there could be some optimization elsewhere - it seems like an easy improvement might be to include some memoization for the regex functions in route-recognizer? I could be overlooking some possible downsides here, and I'm not super familiar with this library. Does this seem like a reasonable suggestion?

I did some testing with a local fork and it seemed to have significant improvements on my test execution time. FWIW if this suggestion gets any traction, I'm happy to do the work to implement it.