Inconsistency in max_startups/shutdowns of a Flow

[p-snft]

The tests added in 896f73b reveal an inconsistency: While shutting down after the optimisation period is not counted, starting up before the optimisation period seems to be counted. In test_maximum_startups, I'd expect the flow to be active in the first and the last time step. Only the startup before the last time step would count.

Originally posted by @p-snft in #1080 (comment)