test(spec): state, multimethods, protocols — fix reify, get-method, hierarchies

Add spec suites for stateful refs (atoms/volatiles/delays/promises), multimethods
(dispatch + hierarchies), and protocols/types/records. Bugs caught and fixed:
- reify with multiple methods registered only the first (stepped i by 2 over
  one-element specs); now collects every method spec
- get-method/methods/remove-method evaluated their arg to the dispatch fn, but a
  multimethod's methods live on its var — now resolve the var (get-method and
  methods are special forms; remove-method fixed). get-method/methods added.
- no global hierarchy: 2-arg derive modified a throwaway, isa? was hardcoded
  false, parents missing, ancestors/descendants returned arrays (so contains?
  failed), and dispatch ignored derive. Now a global hierarchy backs the 1/2-arg
  derive/isa?/parents/ancestors/descendants (returning sets) and multimethod
  dispatch falls back to it.

Corrected a phase5 port assertion that encoded the old isa? bug.
conformance 218/218, jpm test green.

test/integration/ports/phase5-test.janet

@@ -6,9 +6,10 @@
 (print "22: hierarchy...")
 (let [ctx (init)]
   (assert (= true (ct-eval ctx "(map? (make-hierarchy))")) "make-hierarchy returns map")
-  # 2-arity derive/isa? just returns nil/false (no global hierarchy)
+  # 2-arity derive/isa? use the global hierarchy
   (ct-eval ctx "(derive ::square ::shape)")
-  (assert (= false (ct-eval ctx "(isa? ::square ::shape)")) "isa? 2-arity always false"))
+  (assert (= true (ct-eval ctx "(isa? ::square ::shape)")) "isa? 2-arity via global hierarchy")
+  (assert (= true (ct-eval ctx "(isa? ::square ::square)")) "isa? reflexive"))
 (print "  passed")
 
 # 23. Multimethods — basic dispatch