remove turning off autograd by default for torch backend

this used to be required for truly backend-agnostic code in previous keras versions, but seems to be safe to omit now

Author: LarsKue

bayesflow/__init__.py

@@ -32,14 +32,6 @@ def setup():
 
     from bayesflow.utils import logging
 
-    if keras.backend.backend() == "torch":
-        # turn off gradients by default
-        import torch
-
-        torch.autograd.set_grad_enabled(False)
-
-        logging.warning("Disabling gradients by default. Use\nwith torch.enable_grad():\nin custom training loops.")
-
     logging.debug(f"Using backend {keras.backend.backend()!r}")