Improvements to code generation for "foreach"

Specialize the code for the innermost loop to not do any masking
computations for the innermost dimension for the iterations where
we are certainly working on a full vector's worth of data.

This fix improves performance/code quality of "foreach" such that
it's essentially the same as the equivalent "for" loop.

Fixes issue #151.

ctx.h

@@ -161,10 +161,8 @@ public:
     void EndLoop();
 
     /** Indicates that code generation for a 'foreach' or 'foreach_tiled'
-        loop is about to start.  The provided basic block pointer indicates
-        where control flow should go if a 'continue' statement is executed
-        in the loop. */
-    void StartForeach(llvm::BasicBlock *continueTarget);
+        loop is about to start. */
+    void StartForeach();
     void EndForeach();
 
     /** Emit code for a 'break' statement in a loop.  If doCoherenceCheck
@@ -232,6 +230,8 @@ public:
 
     bool InForeachLoop() const;
 
+    void SetContinueTarget(llvm::BasicBlock *bb) { continueTarget = bb; }
+
     /** Step through the code and find label statements; create a basic
         block for each one, so that subsequent calls to
         GetLabeledBasicBlock() return the corresponding basic block. */