switch implementation of heap from being based on heap.sml (heap as described by one of Chris Okasaki's papers) to being based on bin-vec.sml (simple vector storing elements in sorted order); more performant this way for cache reasons

fcore/bin-vec.sml

@@ -12,13 +12,16 @@ sig
   type elem
 
   val empty: elem vector
+  val isEmpty: elem vector -> bool
 
   val sub: elem vector * int -> elem
   val contains: elem * elem vector -> bool
+  val findMin: elem vector -> elem
 
   val findInsPos: elem * elem vector -> int
   val insert: elem vector * elem * int -> elem vector
   val delete: elem vector * elem -> elem vector
+  val deleteMin: elem vector -> elem vector
   val updateAtIdx: elem vector * elem * int -> elem vector
 
   val fromList: elem list -> elem vector
@@ -30,8 +33,23 @@ struct
 
   val empty = Vector.fromList []
 
+  fun isEmpty vec = Vector.length vec = 0
+
+  fun deleteMin vec =
+    if Vector.length vec <= 1 then
+      Vector.fromList []
+    else
+      let
+        val len = Vector.length vec - 2
+        val slice = VectorSlice.slice (vec, 1, SOME len)
+      in
+        VectorSlice.vector slice
+      end
+
   val sub = Vector.sub
 
+  fun findMin vec = Vector.sub (vec, 0)
+
   fun reverseLinearSearch (pos, findNum, vec) =
     if pos < 0 then
       ~1
@@ -183,3 +201,15 @@ structure ValSet =
 
        fun g ({distance = a, ...}: elem, {distance = b, ...}: elem) = a > b
      end)
+
+structure DistVec =
+  MakeBinVec
+    (struct
+       type elem = {distance: int, id: int, comesFrom: int}
+
+       fun l ({distance = a, ...}: elem, {distance = b, ...}: elem) = a < b
+
+       fun eq ({distance = a, ...}: elem, {distance = b, ...}: elem) = a = b
+
+       fun g ({distance = a, ...}: elem, {distance = b, ...}: elem) = a > b
+     end)