Add a bound to the inference tips cache

Small bounds still yield about equal hits and misses.

Further work could determine if storing only the last result
is optimal.

Author: jacobtylerwalls

astroid/inference_tip.py

@@ -6,6 +6,7 @@
 
 from __future__ import annotations
 
+from collections import OrderedDict
 from collections.abc import Generator
 from typing import Any, TypeVar
 
@@ -18,9 +19,9 @@
     TransformFn,
 )
 
-_cache: dict[
+_cache: OrderedDict[
     tuple[InferFn[Any], NodeNG, InferenceContext | None], list[InferenceResult]
-] = {}
+] = OrderedDict()
 
 _CURRENTLY_INFERRING: set[tuple[InferFn[Any], NodeNG]] = set()
 
@@ -61,10 +62,14 @@ def inner(
             _CURRENTLY_INFERRING.add(partial_cache_key)
             try:
                 # May raise UseInferenceDefault
-                result = _cache[func, node, context] = list(func(node, context, **kwargs))
+                result = _cache[func, node, context] = list(
+                    func(node, context, **kwargs)
+                )
             finally:
                 # Remove recursion guard.
                 _CURRENTLY_INFERRING.remove(partial_cache_key)
+                if len(_cache) > 64:
+                    _cache.popitem(last=False)
 
         yield from result