Parallel support for reaktoroblock

[avdudchenko]

With larger thermodatabases such as wateq4f.dat and many blocks (5+) and species (9+) the solving process can become extremely slow (2-3 seconds per jacobian call)

This can be significantly improved by parallelizing reaktoro solver calls.

There are 2 issues, reaktoro it self can not be pickled.
GrayboxModels are executed serially in current implementation.

Proposed approach:

• Create a ReaktoroBlockManager that can be passed into ReaktoroBlock for their aggregation and management
  --This manager will configure reaktoro as normal on main thread, and create a parallel worker that houses reaktoro it self for running solves
  --The manager will configure a single ReaktoroGrayBox model that aggregates all of the sub-reaktoro blocks together

Functionality:
-The ReaktoroBlocks will work as normal when a ReaktoroBlockManager object is not provided
-When ReaktoroBlockManager object is provided, all standard functionality will remain the same, except the user will need to call ReaktoroBlockManager.build() after constructing all Reaktoro blocks on the model, and before running any initialization commands - maybe it will be possible to automate this by catching when user calls to initialize the first ReaktoroBlock.
-The user will ideally will be able to interact with each Reaktoro block directly as before (e.g. initialization calls are still performed on the ReaktoroBlock and not ReaktoroBlockManager, as well as interacting with any options or display functions).