Fix identical omts spawning over themselves

[akrieger]

Summary

Bugfixes "Fix overload of zeds and vehicles with older saves"

Purpose of change

Possibly fixes #67837. One fix led to another bug, which leads to this fix. There's a migration safety issue in mapgen where old saves serialized 'bad data' that was resolved at runtime. The runtime fixup was removed, but the saves were not migrated. This is an attempt to implement that save migration.

Describe the solution

I don't really know the exact terminology, but the laymans answer is things like roads were running mapgen 'over themselves' repeatedly. This resulted in things like dozen car pileups and hundreds of zeds in cities. The patch that introduced this (#67773) implemented logic at worldgen? mapgen? time which prevents these redundant roads being added to a 'predecessors' list. I ported that logic into the savegame deserializer to try to handle this at load time so older saves migrate forward successfully. The only thing I'm unsure of is the base condition of zero to 1 predecessors, but I'm assuming that any predecessor should be allowed compared to a base case of I'm guessing null terrain?

Describe alternatives you've considered

Testing

Created a save in a game compiled just prior to #67773. Debug map revealed and teleported to a city extremely far away. No problem.
Loaded that save on ##67773. Teleported to the same tile. Hundreds of zeds.

Applied my fix. Teleported to the tile. Not hundreds of zeds.

Roads don't have infinite predecessors anymore. A before comparison:

Created a new world. Inspected 4 way intersections in cities. Saw there are still some that have a manhole omt as a predecessor. Manholes appear to not be linear, and adding the LINEAR flag is apparently mutually exclusive with NO_ROTATE, so that's not going to change. I don't know if this means city intersections have a higher chance of double spawns of somethings or not.

Additional context

0.G didn't have this redundant manhole under crossroad stuff. But JSONizing roads is radically different than what was before. No clue how to get there.

[akrieger]

I'm not sure this is a perfect fix. There's still larger pileups happening than should be expected. I'm not sure how far back to go to get to 'true legacy code' to compare against.

[akrieger]

It looks like the behavior is intentional if I'm reading #67223 right, it just seems like spawn rates on roads in town are still elevated. I wonder if just removing REQUIRES_PREDECESSOR from the road json definition would resolve it, but I don't understand the implications of doing that. I will try it though.

[akrieger]

I think I got it. I think the right answer is to not push a linear predecessor if the incoming type is itself linear, unless predecessors is empty. That way a road spawning over a field pushes the field predecessor, if the road is then refined to a more specialized road, it does not push the base road spawn as a predecessor.

[akrieger]

Well I think this approach is plausible. What do people think?