Per #3752, fixes rP7551.
0 A.D. needs knowledge of turns N-1 and N to render things, as it interpolates between the two. If it reaches turn N and doesn't have info on turn N+1, it must freeze (which we've all experience in MP with lag).
To work around this issue, in MP commands are sent not for turn N+1 but N+2, which gives more time for things to get synchronised. The current delay is 2 * turn = 1000ms.
rP8400 originally increased MP turn lengths to increase that delay, but another option is increasing the COMMAND_DELAY.
Increasing the MP turn length is arguably worse, as it:
- makes pathfinding worse
- makes the game more 'lag-spikey', since computations are done less often, but thus then can take more time.
This is all summarised on #3752's comments.
This diff reverts that and instead increases COMMAND_DELAY in MP. This makes turns 200ms in both SP and MP, but MP 'inherent lag' is still higher (though at 800ms instead of 1000ms).
I chose 800 ms because 600ms seems too low (with 200ms ping, client A -> server -> client B will be 400 ms inherent + turn time), and 1000ms seems un-necessarily high. Can easily be changed in the future. It's a 200ms improvement on A24.
SP was also not actually using COMMAND_DELAY, instead just pushing directly. This changes that.