Rework processing of all http requests

[nvborisenko]

Instead of creation new Task, which is continuation of dependent Tasks, we can do the following:

• Put job into queue
• Define 10 consumers which will look into queue

It will allow:

• Less cpu/memory consumption
• No need throttler at all
• We allocate resources exactly as how much we need
• Consumer is long-live thread, less stressing for ThreadPool
• We save http requests order, so real-time reporting is more real