Release v19.59.4

Added

• Cratis:Chronicle:Shutdown:DrainTimeout (environment variable Cratis__Chronicle__Shutdown__DrainTimeout) (default 8 seconds) bounds how long the kernel spends finishing in-flight observer work when it stops; the effective budget is capped below the host shutdown timeout and logged at startup (#4790)

Fixed

• Stopping a kernel silo now waits for in-flight projection catch-up and replay work and writes its buffered MongoDB bulk writes within the shutdown budget, instead of leaving writes that can land after the silo reported it had stopped; writes cut off by the deadline are logged with their observer and partitions and recorded as failed partitions where time allows (#4790)