Clarify Rate Sampling config

[joe-warren]

The documentation for RateSpanSampler, states that it uses a Token Bucket algorithm, and lists the Token Bucket parameters as bucketSize (an Integer) and tokenRate (a Duration).

My reading of the Token Bucket algorithim is that the effective rate limit is bucketSize spans every tokenRate time period (per pod/instance/etc). I think this would be worth explicitly stating?

I also personally feel like the example values (bucketSize = 100, tokenRate = 10.millis) are a little high for trace collection, and a stricter rate limit might make for a better example?