Error when applying the adjoint of a DiagonalStack of TomographicProjector

[Michael-T-McCann]

The following code produces and error

import numpy as np
from scico.linop.radon_astra import TomographicProjector
from scico.linop import DiagonalStack
import scico.numpy as snp

N = 10
H = DiagonalStack([TomographicProjector((N, N), 1.0, N, np.linspace(0, np.pi, N))])
H.T @ snp.zeros(H.output_shape)

NotImplementedError: Transpose rule (for reverse-mode differentiation) for 'custom_vjp_call_jaxpr' not implemented

The error goes away if jit=False is passed to DiagonalStack. But, oddly, I can't reproduce it using a TomographicProjector alone without the stack. May relate to #323.