Solving with complex initialization

When trying to run a basic example:

from diffrax import diffeqsolve, ODETerm, Dopri5
import jax.numpy as jnp
from jax.lax import cond


Array.set_default_backend('jax')
def f(t, y, args):
    return -y

term = ODETerm(f)
solver = Dopri5()
y0 = jnp.array([2., 3.], dtype=complex)
solution = diffeqsolve(term, solver, t0=0, t1=1, dt0=0.1, y0=y0)

I receive this warning: ComplexWarning: Casting complex values to real discards the imaginary part, which goes away if I change the type of y0 to float.

Moreover, when running something else with a complex y0, there was an error TypeError: nextafter does not accept dtype complex64 at position 0. Accepted dtypes at position 0 are subtypes of floating.

Does diffrax not support complex y0s? If so is there any plan to enable this? Although we can convert the complex elements to 2d arrays, it would make the process a lot smoother if we could just use complex elements.

Start with the basic example using diffеqsolve, Dopri5, and a complex y0, then reproduce the ComplexWarning and the nextafter TypeError described in the issue. Trace how complex initial values are handled and verify that supported complex inputs no longer emit the warning or trigger the reported error.