Sync fixes

[seva-ask]

This pull request fixes rare case of ignoring listener of aggregated requests. Please check if my synchronization variant is right and doesn't break anything, but in my project it works.
Problem: if you execute two same requests in very close time, for example from different threads, they are aggregated. But it can happen that listeners were taken from map, then sent to request notifier by progress manager. And before removing from map request processor will take listeners set from map and add listener to it. So that listener will never be called because notification was executed alreafy.

[seva-ask]

Please consider releasing new build to maven after reviewing two my pull requests.
And it might be interesting for you to take a look at this project: https://github.com/ZuZuK/base-library
Sorry, It has lack of documentation now because it is under development.
My friend and I are working on concept of aggregating requests and calling listeners only after all request chain is executed or failed.
Also we want to make caches faster using https://github.com/RuedigerMoeller/fast-serialization
And we want to make transparent for developer such behaviour as: if we have cached values, then show them, then reload and update data from network, and then refresh screen.
If you are interested in, we can discuss it and answer any questions.