Always cache resolvers

After getting access to a second dataset which is a lot more
cache-friendly than the original (while also being a real-world
sample), I think it makes sense to cache the "fast" resolvers as well:
going back to the original
commit (b45380d) an S3Fifo(5000) was
estimated to take about 3.5MB when full (0.5MB of cache metadata, and
3MB of actual cached data), so an S3Fifo(2000) would be about 1.5MB or
thereabouts which is not nothing... but is pretty much nothing
compared to the RSS impact of the advanced parsers.

Fixes #302

Author: masklinn

src/ua_parser/__init__.py

@@ -68,14 +68,7 @@
 if importlib.util.find_spec("ua_parser_rs"):
     from .regex import Resolver as RegexResolver
 BestAvailableResolver: _ResolverCtor = next(
-    filter(
-        None,
-        (
-            RegexResolver,
-            Re2Resolver,
-            lambda m: CachingResolver(BasicResolver(m), Cache(2000)),
-        ),
-    )
+    filter(None, (RegexResolver, Re2Resolver, BasicResolver))
 )
 
 
@@ -97,7 +90,7 @@ def from_matchers(cls, m: Matchers, /) -> Parser:
         stack.
 
         """
-        return cls(BestAvailableResolver(m))
+        return cls(CachingResolver(BestAvailableResolver(m), Cache(2000)))
 
     def __init__(self, resolver: Resolver) -> None:
         self.resolver = resolver