Revert lax_numpy.asclose() behavior to work with lists again. (google…

…#2059)

This should be revisited to fix the issue originally addressed in google#2051.

jax/numpy/lax_numpy.py

@@ -314,12 +314,12 @@ def update_numpydoc(docstr, fun, op):
 
 def _wraps(fun, update_doc=True, lax_description=""):
   """Like functools.wraps but works with numpy.ufuncs.
-     It is important that when wrapping numpy functions the parameters names 
+     It is important that when wrapping numpy functions the parameters names
      in the original function and in the JAX version are the same
     Parameters:
       fun: The function being wrapped
       update_doc: whether to transform the numpy docstring to remove references of
-      parameters that are supported by the numpy version but not the JAX version. 
+      parameters that are supported by the numpy version but not the JAX version.
       If False, include the numpy docstring verbatim.
   """
   def wrap(op):
@@ -1021,7 +1021,7 @@ def moveaxis(a, source, destination):
 
 @_wraps(onp.isclose)
 def isclose(a, b, rtol=1e-05, atol=1e-08):
-  a, b = _promote_args("isclose", a, b)
+  a, b = _promote_args("isclose", asarray(a), asarray(b))
   dtype = _dtype(a)
   if issubdtype(dtype, inexact):
     if issubdtype(dtype, complexfloating):