Stuck or very Slow check is_linear_expression in LinearForm and BilinearForm

[YzzIzzY]

is_linear_expression checks whether the expression given in a LinearForm (or a BilinearForm resp.ly) instance is actually linear (bi-linear resp.ly). For some expressions, known to be linear, the check is really slow or even stuck in a loop. This might be due to the current version used of sympy in sympde.
As an example, you may import Laplacian from psydac.examples.poisson_2d_mapping and TargetMapping from sympde.topology.analytical_mapping and run

params   = dict(c1 = 0, c2 = 0, k = 0.3, D = 0.2)
mapping  = TargetMapping('F', **params)

lapl  = Laplacian(mapping)
s, t  = mapping.logical_coordinates
x, y  = mapping.expressions
k     = params['k']
D     = params['D']
kx    = 2 * pi / (1 - k + D)
ky    = 2 * pi / (1 + k)
phi = (1 - s**8) * sin(kx * (x - 0.5)) * cos(ky * y)
rho = - lapl(phi)

logical_domain = Square('Omega', bounds1 = (0, 1), bounds2 = (0, 2 * pi))
domain  = mapping(logical_domain)
V0 = ScalarFunctionSpace('V0', domain)

u0, v0 = elements_of(V0, names = 'u0, v0')
rhs = LinearForm(v0, integral(domain, rho * v0))

[yguclu]

I am reopening this issue, as the problem still shows up in certain situations (e.g. in the PSYDAC example poisson_3d_target_torus.py, as discussed here). PR #187 made significant improvements toward solving this issue, but we need a more structural and solid solution. In particular, I think a new API design might be essential