[Dev] Allow exa vectors

[jbcaillau]

Parse x(t) (vs. x[i](t)) to parse to sth like [x[i] for i in 1:n] since x[:, j] is not permitted in ExaModels

[jbcaillau]

• add subsx to allow that after x[i](t) substitution (the converse could work)
• test
• try with quantum control example using an external function working

see also:

• extend [:] operator, same for +, *... on ExaModels.variable
• X = [x[i, j] for i in 1:n, j in 0:N] # in X, j is in 1:N+1

[jbcaillau]

• call functions (overload of *...) that return exa values (e.g. arrays of exa vars, cf. above)
• use aliases like: Qt = reshape(q(t), n, 2n)
• complex values with real + imag on separate columns -> better: use tensor with extra dim of length one for real, length two for complex
• actually tensorise everything to be matrix ready (need for more?): R^n is M(n, 1, R); allow for, more generally, M(n, m, K) with K = R or C
• set x = variable(c, n, nn, dim_K, 0:N) and X = [x[i, ii, k, j] for i in 1:n, ii in 1:nn, k in dim_K, j in 0:N], then use X for all computations
• import Base: * + redefinition for product, etc.
• check w = Vector{ExaModels.AbstractNode}(undef, 4) to store results

NB. use systematically a modeller to define the optimisation problem, even for "functional" case?