Update effective_rate calculation

[ericmustin]

Summary

This PR Updates the Effective Rate calculation of the Rule Sampler. According to Spec, the effective rate should be the average of the last 2 time windows, each time window 1s each.

Notes

I mainly tried to mimic the dd-trace-py https://github.com/DataDog/dd-trace-py/blob/8c450b56ba87cd8cab916bb579a01837c35b0c01/ddtrace/internal/rate_limiter.py#L7 approach. We have some existing tests that mock a >1s time period and they all still pass, but if there's conditions we should test here I can try to add them.

[brettlangdon]

Python has to update since they have it slightly wrong.

We want to change the formula from:

rate = (previous_rate + current_rate) / 2

to

rate  = (previous_window_sampled_traces + current_window_sampled_traces) / (previous_window_total_traces + current_window_total_traces)

It'll give us more consistently accurate values over the two windows.

[brettlangdon]

did we need to adjust any tests for this?

[ericmustin]

@brettlangdon Updated the specs to have tests more specific to the behavior we've added, as our previous tests were a bit too broad to ensure we can prevent a regression here.