Spatial ssa support

[TorkelE]

I have added support for spatial jumps. Creating and solving a problem is relatively easy:

using Catalyst, Graphs, JumpProcesses

# Make model.
SIR_system = @reaction_network begin
    α, S + I --> 2I
    β, I --> R
end
trS = @transport_reaction dS S
trI = @transport_reaction dI I
lattice = Graphs.grid([2, 2])
lrs = LatticeReactionSystem(SIR_system, [trS, trI], lattice)

# Parameter values and initial conditions.
u0 = [:S => [999, 1000, 1000, 1000], :I => [1, 0, 0, 0], :R => 0]
pV = [:α => 0.1 / 1000, :β => 0.01]
pE = [:dS => [0.1, 0.1, 0.1, 0.2], :dI => [0.01, 0.01, 0.01, 0.02]]

# Create the problem.
dprob = DiscreteProblem(lrs, u0, (0.0,100.0), (pV,pE))
jprob = JumpProblem(lrs, dprob, NSM())
sol = solve(jprob, SSAStepper())

Current limitations:

• Cartesian grids are not supported, only Graphs.
• Of the 5 forms of the hopping rates, we now convert to the most general one (D_{s,i,j}). Once this is merged, I will introduce a routine which detects which is the most compact form that can be provided, and uses that one.
• Currently, the constant rate jumps in each compartment must have the same parameter values. Once this is supported in JumpProcesses I will add support for it here as well.

[TorkelE]

Rebased on #644

[TorkelE]

Superseded by #737