JS: Update featurization for absent features optimization

Absent features are now represented implicitly by the absence of a row in the `tokenFeatures` relation, rather than explicitly by an empty string. This leads to improved runtime performance. To enable this implicit representation, we pass the set of supported token features to the `scoreEndpoints` HOP. Requires CodeQL CLI v2.7.4.

Author: henrymercer

javascript/ql/experimental/adaptivethreatmodeling/lib/experimental/adaptivethreatmodeling/EndpointFeatures.qll

@@ -261,7 +261,7 @@ private module AccessPaths {
 }
 
 /** Get a name of a supported generic token-based feature. */
-private string getASupportedFeatureName() {
+string getASupportedFeatureName() {
  result =
  [
  "enclosingFunctionName", "calleeName", "receiverName", "argumentIndex", "calleeApiName",
@@ -277,12 +277,5 @@ private string getASupportedFeatureName() {
  */
 predicate tokenFeatures(DataFlow::Node endpoint, string featureName, string featureValue) {
  ModelScoring::endpoints(endpoint) and
- (
- if strictcount(getTokenFeature(endpoint, featureName)) = 1
- then featureValue = getTokenFeature(endpoint, featureName)
- else (
- // Performance note: this is a Cartesian product between all endpoints and feature names.
- featureValue = "" and featureName = getASupportedFeatureName()
- )
- )
+ featureValue = getTokenFeature(endpoint, featureName)
 }

javascript/ql/experimental/adaptivethreatmodeling/lib/experimental/adaptivethreatmodeling/EndpointScoring.qll

@@ -95,7 +95,8 @@ module ModelScoring {
  }
 
  predicate endpointScores(DataFlow::Node endpoint, int encodedEndpointType, float score) =
- scoreEndpoints(endpoints/1, requestedEndpointTypes/0, relevantTokenFeatures/3,
+ scoreEndpoints(endpoints/1, relevantTokenFeatures/3,
+ EndpointFeatures::getASupportedFeatureName/0, requestedEndpointTypes/0,
  getACompatibleModelChecksum/0)(endpoint, encodedEndpointType, score)
 }