Ensured that dtype subclasses are hashable

Python data model requires a class to implement both `__eq__` *and* `__hash__`
to be considered hashable. If a class only implements `__eq__`, it gets an
auto-generated `__hash__ = None`, which makes it non-hashable.

So, as things stand, `tl.pointer_type` and `tl.block_type` instances fail
to hash. The fix is to define __hash__ explicitly in the class body, as
suggested in [*]:

> If a class that overrides __eq__() needs to retain the implementation of
> __hash__() from a parent class, the interpreter must be told this explicitly
> by setting __hash__ = <ParentClass>.__hash__.

[*]: https://docs.python.org/3/reference/datamodel.html#object.__hash__

python/test/unit/language/test_core.py

@@ -513,6 +513,14 @@ def test_dtype_codegen():
         assert repr(eval(full_name)) == full_name
 
 
+@pytest.mark.parameterize("dtype", [
+    tl.pointer_type(tl.int8),
+    tl.block_type(tl.int8, [42])
+])
+def test_dtype_is_hashable(dtype):
+    hash(dtype)
+
+
 # ---------------
 # test binary ops
 # ---------------

python/triton/language/core.py

@@ -594,6 +594,8 @@ def __eq__(self, other: pointer_type) -> bool:
     def __ne__(self, other: pointer_type) -> bool:
         return not self.__eq__(other)
 
+    __hash__ = dtype.__hash__
+
     @property
     def scalar(self):
         return self
@@ -646,6 +648,8 @@ def __eq__(self, other: block_type) -> bool:
     def __ne__(self, other: block_type) -> bool:
         return not self.__eq__(other)
 
+    __hash__ = dtype.__hash__
+
     @property
     def scalar(self):
         return self.element_ty