Monte-Carlo: convergence diagnostics, variance reduction, parallel replications

[derrynknife]

Availability/MTTF simulations now carry seeds, standard errors and confidence bands. The remaining Monte-Carlo hardening tail:

• Convergence diagnostics — e.g. stop replications once the CI half-width is below a tolerance.
• Variance reduction — common random numbers / antithetic variates.
• Parallel replications across cores.

Estimate

Effort: L (~1–2 weeks) · Value: Med

Three separate sub-features; parallel replications in particular is fiddly given the global-RNG seeding model. The simulator already has seeds, standard errors and Wilson bands, so this is diminishing-returns polish — worth deferring until a user actually hits a simulation that's too slow or too noisy.