Restart for ALMOST_SOLVED integer solutions

[Wikunia]

Currently if almost locally solved integer solutions aren't allowed but almost locally solved is one restart is done to maybe obtain a good solved integer solution. This restart should probably be an option.

• removing restarts
• maybe try restart but if not better accept it nevertheless?

[ccoffrin]

@Wikunia I was having a look at the current implementation of state_is_optimal and given that allow_almost is in there it made we wonder if this issue has been addressed or if there is something outstanding to investigate?

[Wikunia]

I think the question is whether we always allow "almost solved" or whether it makes sense to only allow "locally solved". If the latter one might want to run the problem with a different seed to obtain an actual "locally solved"

[ccoffrin]

Got it. I will leave this as an open question as some investigation is needed to understand the possible impact.

My hunch is that almost solved models have numerical issues and the almost property will be stable across multiple solutions, but this should be verified.