Honest scenario / batch runner

Every scenario in a batch/sensitivity run gets its own independent, real solve — a scenario that fails to converge is reported FAILED and excluded from governing-case comparisons, never silently folded in alongside the ones that actually solved as if it had succeeded too.

• You're running a scenario tree, tornado-sensitivity sweep, or multi-contingency flare-header rating and need to trust that a 'governing case' really is the worst CONVERGED case.
• A batch result set looks suspiciously complete (every scenario shows a number) and you want to confirm none of those numbers came from a failed solve.
• You're building a new multi-run feature (a sweep, a scenario comparator) and need the pattern for not letting one bad run contaminate the set.

Each scenario solve is independently gated by the same convergence checks a single solve gets, and a FAILED scenario is excluded (not zero-filled or carried-forward) from any aggregate like 'governing case' or 'worst backpressure' — this is also the mechanism behind a dedicated blocking note that fires when excluding a failed scenario would make the remaining governing case non-conservative.