Enhance calculation of tokens number in response based on request's max_tokens parameter - histogram should be adapted to the max_tokens

[mayabar]

[mayabar]

The histogram used in the initial solution (equally sized buckets) will not give a right distribution when maxTokens is higher than 100. In this case we will use 3 first buckets of size 20, the 5th bucket of size 20, the last bucket of size 1, but the 4th bucket will contain all other values.
For example:
maxToken = 50, buckets: 1-9, 10-19, 20-29, 30-39, 40-49, 50
maxToken = 500, buckets: 1-19, 20-39, 40-59, 60-479, 480-499, 500