[Backend Tester] Don't run fp64 model tests (#13982)

They add to a lot of failures and double precision isn't commonly used.
We can look at re-enabling these in the future. Ideally all backends
would just not delegate if they can't run it.

Author: GregoryComer

backends/test/suite/models/__init__.py

@@ -22,7 +22,6 @@
 DTYPES: list[torch.dtype] = [
     torch.float16,
     torch.float32,
-    torch.float64,
 ]