Renumbering sandbox policy actions to fix bug found by VC++'s /analyze.

Warning was:
src\sandbox\win\src\policy_engine_processor.cc(81) : warning C6313:
Incorrect operator:  zero-valued flag cannot be tested with bitwise-and.

This change should have no effect because the kStopOnErrors policy is
never requested.

BUG=427616

Review URL: https://codereview.chromium.org/789543004

Cr-Commit-Position: refs/heads/master@{#307988}

sandbox/win/src/policy_engine_processor.h

@@ -63,18 +63,18 @@ enum PolicyResult {
 };
 
 // Policy evaluation flags
-// TODO(cpu): implement the options 0 & 4.
+// TODO(cpu): implement the options kStopOnErrors & kRankedEval.
 //
 // Stop evaluating as soon as an error is encountered.
-const uint32 kStopOnErrors = 0;
+const uint32 kStopOnErrors = 1;
 // Ignore all non fatal opcode evaluation errors.
-const uint32 kIgnoreErrors = 1;
+const uint32 kIgnoreErrors = 2;
 // Short-circuit evaluation: Only evaluate until opcode group that
 // evaluated to true has been found.
-const uint32 kShortEval = 2;
+const uint32 kShortEval = 4;
 // Discussed briefly at the policy design meeting. It will evaluate
 // all rules and then return the 'best' rule that evaluated true.
-const uint32 kRankedEval = 4;
+const uint32 kRankedEval = 8;
 
 // This class evaluates a policy-opcode stream given the memory where the
 // opcodes are and an input 'parameter set'.