Use compare_by_identity on some runtime caches

[rmosolgo]

Some hashes in GraphQL-Ruby are by-type caches of things. Those will always have the same objects (Classes) as keys, so we could use compare_by_identity for them (suggested in #4431 (comment))

Initial results are good:

  Warming up --------------------------------------
  Querying for 1000 objects
-                           4.333  (± 0.0%) i/s -     22.000  in   5.082679s
+                           4.848  (± 0.0%) i/s -     25.000  in   5.158312s

  # ... 
  ==================================
       TOTAL    (pct)     SAMPLES    (pct)     FRAME
-      26656   (9.4%)       26656   (9.4%)     Kernel#hash
+      17766   (2.7%)       17766   (2.7%)     Kernel#hash


-      26300   (9.3%)       26300   (9.3%)     Kernel#class
+      68174  (10.4%)       68174  (10.4%)     Kernel#class

TODO:

• Address some usages of schema.after_lazy to use this runtime method instead (perhaps by adding query. or context.after_lazy?) which uses a better cache