fix: count conflict resolutions as steps in the `lia` model search

This PR makes `liaSteps` also bound the cases enumerated by Cooper and disequality splits in `lia`, so that a long enumeration stops at the threshold instead of producing a proof the kernel rejects.

The step counter only ticked when a variable was processed. The next case of a split can be refuted as soon as it is asserted, so an enumeration of thousands of cases never reached the check. Conflict resolutions now count as steps; the search is interrupted only below a decision, where `resetDecisionStack` discards the pending conflict together with the cases.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>