AI great prophet usage tweaks

[OptimizedForDensity]

Allows the AI to use great prophets to build holy sites if they can't found/spread/enhance a religion, which prevents some civs from accumulating prophets.

Also fixes a bug where some great prophets may idle in foreign lands. The inquisitor AI has had the same fix applied to it. The below code is from the current repo version.

val destination = cities.getTiles().asSequence()
            .filterNot { unit.getTile().owningCity == it.owningCity } // to prevent the ai from moving around randomly
            .filter { unit.movement.canMoveTo(it) }
            .sortedBy { it.aerialDistanceTo(unit.currentTile) }
            .firstOrNull { unit.movement.canReach(it) } ?: return

unit.movement.headTowards(destination)

You can see that destination tries to find the closest tile to cities (which is always just a single city). The first line gets all the city's tiles then the filterNot tries to remove all tiles that belong to the city whose tiles the unit is standing in. This was implemented to fix a behavior where missionaries would move around randomly, however due to the problem described in #7039, destination would always be null and the prophet would never spread religion.

Changing the filter predicate to unit.movement.canMoveTo(it) || it == unit.getTile() fixes both the random movement behavior and always allows for the missionary to spread religion to the tile it's on. canMoveTo(tileInfo) could be changed to always return true if the tile argument is the same one the unit is standing on, but given how complex pathfinding I don't know if this would break anything.

[Azzurite]

which is always just a single city

I would actually assume that there are edge cases where it's not a "always" a single city - our pathing algorithm could cause the prophet to take a route that temporarily brings it closer to another city, and then it would switch. But it's good enough, just musing :D

canMoveTo(tileInfo) could be changed to always return true

It should definitely. There are a couple places where the current tile is being checked already, exactly like you do here, probably also to work around this bug.