JaxLinearOperator: Fix error in adjoint_times when supplying manually implemented transpose function

The apply-method of JaxLinearOperator expects func_T to return a tuple of values because jax.linear_transform does so. If the user supplies a manually implemented transpose function implementation which returns just a plain value array, this assumption is broken, causing a processing error in the apply method.

To fix this, this patch wraps manually supplied transpose function implementations in a lambda function that encapsulates their result in a tuple.

src/operators/jax_operator.py

@@ -122,7 +122,8 @@ class JaxLinearOperator(LinearOperator):
     func_T : callable
         The jax function that implements the transposed action of the operator.
         If None, jax computes the adjoint. Note that this is *not* the adjoint
-        action. Default: None.
+        action. Needs to return a tuple with the result as the first entry.
+        Default: None.
 
     domain_dtype:
         Needs to be set if `func_transposed` is None. Otherwise it does not have