Multithreaded loading support

[IvanSavenko]

Early WIP.

Goals:

• Allow loading of game in multiple threads, improving game startup and server startup times
• Allow general tasks support, in any part of our code, including running tasks from inside tasks - e.g. run parallel load of each handler inside parallel load of each mod

Changes:

• Moved ThreadPool class from RMG to root directory of library
• Removed old version of ThreadPool that was used in VCAI
• Simplified usage of ThreadPool and removed excessive waits / locks
• Added parallel loading of Json's in mod loading
• Made thread pool global instead of creating it in place
• Allow running tasks from inside tasks without spawning dozens of threads

Progress:

• allow multithreaded access to IdentifiersStorage during loading
• parallelize ObjectTemplate::readTxt
• parallelize CModHandler::loadMod per mod
• parallelize CModHandler::loadMod per handler
• parallelize CModHandler::loadMod per object
• parallelize CModHandler::preloadModData per handler
• parallelize CModHandler::preloadModData per mod
• parallelize CModHandler::calculateModChecksum per mod
• parallelize or remove CModHandler::validateTranslations

Regressions:

• add async schema loading or load all schemas before validation stage

Possible improvements:

• check profiler for any remaining issue, such as delays caused by wait/sleep
• consider some optimization to logging API or remove excessive logging during parallel stage if necessary

[nullkiller]

Are you sure we should keep static instance in header. Also we have TBB anyway with its own thread pool. So maybe it is less convenient to use it, I do not know

[nullkiller]

Probably TBB is more complex.

[IvanSavenko]

Static inside inline function should work as intended, so that part is OK.

Thread pool from TBB - will look into it.
If they have persistent, preconfigured thread pool - then probably makes sense to use it. But probably an overkill for our needs.

I am aiming for minimal overhead / minimal configuration to a point where you can parallelize operation that takes 1-2 ms. Meanwhile TBB:

CAUTION:
Typically a loop needs to take at least a million clock cycles to make it worth using parallel_for. For example, a loop that takes at least 500 microseconds on a 2 GHz processor might benefit from parallel_for.