Quickstart AI Battle

[wichern]

This change allows to quick-start into an AI-only game.

Example:
./bin/s25client -m share/s25rttr/RTTR/MAPS/NEW/Bagel5.SWD --ai aijh --ai aijh --ai dummy

Will result in

Implementation:
This change adds a new program option (--ai) and extends QuickStartGame().
In order to have the requested AIs ready in dskGameLobby, the AI configuration is stored in GameClient.
In GameClient::GameLoaded() we call ToggleHumanAIPlayer() to turn the human player into an AI as well.

Notes:
There are no further limitations. The user can still add buildings and roads.

[Flamefire]

Good idea, well done overall, just a few bits for polishing 👍

[wichern]

@Flamefire
Thank you very much for your review!
I tried to fix all your findings.

Tomorrow, I'll check if it is possible to save and load an ai battle.

[wichern]

Starting an AI-Battle from a savegame seems to work.

The provided AI players (--ai) have no effect in that case, except when none is given at all. In the latter case, player 0 will be played by the human player.

[Flamefire]

Looking good so far, just a few minor improvements, especially making stuff const correct/const-where-possible,
Just a remark: Short-ish single-line statements after if/for/... don't need braces. Indentation (enforced via clang-format) already take care of potential bugs. No need to remove them as they don't hurt, just might save you some work.

There is still a bug though now that I'm thinking about it: aiBattlePlayers_ is never reset, this means after the first such quickstart every game will be considered an AI battle. I guess this reset (clear) can be best done in GameClient::Stop

[wichern]

I did not see an existing UT for QuickStartGame(). How should I handle missing code coverage? Would you like me to create a UT for it?

[Flamefire]

I did not see an existing UT for QuickStartGame(). How should I handle missing code coverage? Would you like me to create a UT for it?

As that method is not critical I'm OK with missing coverage for that, as we have the review at least. If you can come up with a test then feel free. However I imagine that this would be very complicated due to the client, server, windowmanager and desktop being involved (so it's rather an integration test). Hence the idea to factor out the AI name parsing to have at least that tested, and skip the rest which "should work" after inspection ;-)