Add of distributed algorithm for the reference table creation in abc model choice. Documentation for abc model choice methods.

change the stop criteration of a ContinuousTimeModel simulation.
Minor fix  in ContinuousTimeModel for the euclidean distance automata
tests.
Add of a test for stop criteria simulation of ContinousTimeModel
(unit/simulation_stop_criteria).

All tests passed.

Fix of distributed abc smc tests + add of a test.
Preparation for the new feature: add a stop criteria in the simulation.
All tests passed.

It computes the euclidean distance and finish the simulation if the
distance is above the tolerance epsilon.
I ha to change some function signature of types in core/model.jl but it
was minor changes.
I add a new test + change some others related to automata.
A minor rewriting core/model for better readability

The solution is to use FunctionWrappers.jl which provides a high
interface for C function pointers.

All tests passed.

update_state functions but performance gets worst.
With bench/pkg/abstract_arrays.jl i've highligthed each time a function
is in a collection performance gets worst.

implemetend in the package.
A lot of files has been modified because some of the internal syntax has
changed. But the top level methods should still work.

Two tests about the plots of synchronized oscillatory simulations have
been added.

All tests passed.

Julia shouldn't crash but rather raise an error about the existence
of a function generated by metaprogramming. I didn't manage to isolate
the segfault withtout the package.
To overcome the issue, I add another level of multiple dispatch/abstract
type for synchronized models.
Test of the euclidean distance automaton works.

Meta code generation for next_state! and simulation of synchronized trajectories. bench/pkg/euclidean_distance*.jl works but segfault with test/automata/euclidean*

This commit groups the change operated to the creation of models and
simulate function of a ContinuousTimeModel.
The general idea is to create a concrete type and a simulate function
 per model creation by metaprogramming.
- Now, ContinuousTimeModel is an abstract type. Each creation of a model
defines a concrete type T <: ContinuousTimeModel by meta programming.
- f! and isabsorbing ContinuousTimeModel fields are Symbols.
- simulate(::ContinuousTimeModel) is run by multiple dispatch, according
to the type of the model.

Can't run the whole tests for now but unit/simulate_available_models.jl
runs properly (i've updated the list of models in this commit), and I've
manually checked in the repl that simulations run correctly (distributed
/ plots).

edge_candidates on top-level of next_state! function.