[frontend][typing] Adding typing & small fix for autotuner.py (#5459)

Hello everyone!

# Context

I've been working on improving the compatibility of torch.compile with
user-defined Triton kernels and recently ran into a type checking (and
potential correctness) issue
[here](https://github.com/pytorch/pytorch/pull/142207/files#diff-f5fa7d0e418e91c63fa56d577a92a294c87e19318a3c8b3736ac4254eaa51db9R1054).

This doesn't type the entire file, but it fixes several typing issues in
autotuner.py. There is a plan to revisit this file and other user-facing
portions of the API---but since this catches a potential bug I thought
it would be worth filing this sooner rather than later.

# Testing 

I ran the pre-commit hooks and `python3 -m python/test/unit`. 

I ran mypy with the following command:
```bash
mypy python/triton/runtime/autotuner.py | grep autotune
```
and the python unit tests.

Author: SamGinzburg

python/triton/runtime/autotuner.py

@@ -4,7 +4,7 @@
 import os
 import time
 import inspect
-from typing import Dict
+from typing import Dict, Tuple, List, Optional
 
 from .jit import KernelInterface
 from .errors import OutOfResources, PTXASError
@@ -23,7 +23,7 @@ def __init__(
         restore_value,
         pre_hook=None,
         post_hook=None,
-        prune_configs_by: Dict = None,
+        prune_configs_by: Optional[Dict] = None,
         warmup=None,
         rep=None,
         use_cuda_graph=False,
@@ -40,7 +40,7 @@ def __init__(
         else:
             self.configs = configs
         self.keys = key
-        self.cache = {}
+        self.cache: Dict[Tuple, Config] = {}
         self.arg_names = arg_names
 
         # Reset to zero or restore values
@@ -211,14 +211,18 @@ def run(self, *args, **kwargs):
         self.nargs = None
         return ret
 
-    def prune_configs(self, kwargs):
+    def prune_configs(self, kwargs: Dict) -> List[Config]:
         pruned_configs = self.configs
         if self.early_config_prune:
             pruned_configs = self.early_config_prune(self.configs, self.nargs, **kwargs)
         if self.perf_model:
             top_k = self.configs_top_k
             if isinstance(top_k, float) and top_k <= 1.0:
                 top_k = int(len(self.configs) * top_k)
+            elif not isinstance(top_k, int):
+                # Slice index must be an integer
+                raise TypeError(f"Error while pruning configs, top_k must be either 1) a float <= 1.0 or 2) an int")
+
             if len(pruned_configs) > top_k:
                 est_timing = {
                     config: self.perf_model(