RandomizedWeightedBalancer chooses the first upstream with higher probability than others

[astsiapanay]

Supposing we have 10 upstreams with equal weight = 1.

The current algo selects a random value from the range [0, upperBound] where upperBound in inclusive
calculates prefixSums for each upstream:

1, 2, 3, 4, 5, 6, 7, 8, 9, 10

Later the algo finds the first index at prefixSums where

prefixSum[index] >= randomValue

where index is an index of the upstream to be selected

The problem is that all upstreams have equal ranges, the random value may be in except for the 1st upstream

[0,1],[2,2],[3,3],...,[10,10]

So the 1st upstream have in twice more possibility to be chosen than others.