Various performance-related improvements

[aaronbembenek]

• Short-circuit rule evaluation: eagerly check if the head of a rule has been derived already and, if so, move on to the next tuple combination.
• Fix reordering of rules, so that delta atoms can always be placed first (if desired).
• Change default settings:
  • codegenSplitOnSmt is now false by default
  • Eager evaluation uses the user-provided smtTaskSize parameter (used to always be 1)
• Add ability to collect statistics on the amount of work done per rule.