CogVM source as per VMMaker.oscog-eem.3382

Add primitive 150, BoxedFloat64>>primitiveCosine and primitive 560,
SmallFloat64>>primitiveCosine.
Make all of BoxedFloat64 primitives 51 through 58, primitiveArctan,
primitiveExp, primitiveExponent, primitiveFractionalPart, primitiveLogN,
primitiveSine, primitiveSquareRoot, primitiveTimesTwoPower,
primitiveTruncated BoxedFloat64 only, no longer checking but insisting
on the receiver being a BoxedFloat64.

Increases performance, e.g. from 898 nanoseconds per invocation to 597
nanoseconds on a 2021 Apple M1 Max MPB.

To implement the above add noFailFloatValueOf: to the object memories.
Discard dbgFloatValueOf: since all its invocations can be replaced with
noFailFloatValueOf:.

src/spur32.cog.lowcode/cointerp.h

@@ -1,5 +1,5 @@
 /* Automatically generated by
-	CCodeGeneratorGlobalStructure VMMaker.oscog-eem.3380 uuid: c077b429-f897-47ad-92b4-563c161255f3
+	CCodeGeneratorGlobalStructure VMMaker.oscog-eem.3382 uuid: 936441c0-8b36-4f43-a893-4c90667b5b58
  */
 
 
@@ -137,6 +137,7 @@ extern sqInt headerIndicatesAlternateBytecodeSet(sqInt methodHeader);
 extern sqInt instantiateClassindexableSize(sqInt classObj, usqInt nElements);
 extern sqInt isIntegerValue(sqInt intValue);
 extern sqInt isMarked(sqInt objOop);
+extern double noFailFloatValueOf(sqInt aFloatOop);
 extern usqInt smallObjectBytesForSlots(sqInt numSlots);
 extern sqInt remember(sqInt objOop);
 extern sqInt addressCouldBeObj(sqInt address);