* Make separate libs Common/ZooKeeper and Common/ConfigProcessor
renamed: dbms/src/Common/StringUtils.h -> dbms/src/Common/StringUtils/StringUtils.h
* Fix space
* Fix data in system.build_options
* merge
* merge
- We add BackgroundSchedulePool which can execute a function at a specific point in time. Basically all tasks are added in a queue and precessed by worker threads.
- The most important difference between this and BackgroundProcessingPool is that we have the guarantee that the same function is not executed from many workers in the same time
- Each of the following classes instead starting a thread will register a task in BackgroundSchedulePool and when they need to run will call schedule or scheduleAfter(duration) functions
This commit is moving all threads created by ReplicatedMergeTree to BackgroundSchedulePool tasks
NOTE: I did a minimum number of changes to be much simple to review the code
The tryGet() operation creates a 1MB buffer on stack. This may or
may not work depending on the default stack size for threads,
whether the stack protector is enabled or not, recursion depth,
and the actual value size.
This is probably going to slow down some ZK operations, but I don't
see how else this could work reliably with the existing API.
std::random_shuffle() may or may not use std::rand() which isn't
seeded, and since configuration parsing runs predictably as one
of the first things after server startup, the list of ZK nodes
is shuffled the same way on all replicas.
