As far as I know type hints don't affect Python's performance, because they're optional and not yet used by the runtime.
In any case, JavaScript is dynamically/unityped, so its performance is unlikely to provide a reason for type annotations. The real problem here is that PyPy is still maturing, and can't be used by many people because of legacy C extensions.
An they're unlikely ever to be used by the runtime, because what hints there are are far too vague. Saying something is class X is useless when you don't know what attributes that class has (because they can be dynamically added and removed) nor what types those attributes are.
The level if information modern Python tools like those in PyCharm and JEDI can reason through to and discover about runtime behaviour is actually pretty impressive. I don't think it's going to pose a huge challenge for these kinds if tools to find. Dealing with ambiguous cases is a UI problem, too much info, not enough space.
They don't affect performance, but I have to agree they really don't look nice to read or write, which is unfortunately the antithesis of Python's philosophy.
In any case, JavaScript is dynamically/unityped, so its performance is unlikely to provide a reason for type annotations. The real problem here is that PyPy is still maturing, and can't be used by many people because of legacy C extensions.