ag_if should auto-cast R atomics to the correct dtype.

[t-kalinowski]

library(tensorflow)
f <- function(x) if(x > 0) x else 0
ff <- tf_function(f)
#> Loaded Tensorflow version 2.8.0
ff(as_tensor(array(1)))
#> Error in py_call_impl(callable, dots$args, dots$keywords): RuntimeError: Evaluation error: TypeError: true_fn and false_fn arguments to tf.cond must have the same number, type, and overall structure of return values.
#> 
#> true_fn output: {'returned': <tf.Tensor 'cond/Identity:0' shape=(1,) dtype=float64>}
#> false_fn output: {'returned': <tf.Tensor 'cond/Identity:0' shape=() dtype=float32>}
#> 
#> Error details:
#> Tensor("cond/Identity:0", shape=(1,), dtype=float64) and Tensor("cond/Identity:0", shape=(), dtype=float32) have different types
#> .

^{Created on 2022-03-10 by the reprex package (v2.0.1)}