faction camp: menial labor uses the zone autosort feature

[mlangsdorf]

Summary

SUMMARY: Features "faction camp: menial labor uses the zone autosort feature"

Purpose of change

Fixes #25397
Fixes #28880
Fix faction camp menial labor sorting by making it use the existing zone autosort feature. As a side effect, add the infrastructure for NPCs performing player activities on an ongoing basis.

Describe the solution

Add a function to clzones to determine if there are any loot zones nearby.

Alter the npc *_destination functions so it is clear that deal with omt destinations.

Add infrastructure to npcmove to let friendly NPCs perform player activities using the NPC_MISSION_ACTIVITY mission. NPCs gain bonus moves from elapsed time during npc::on_load() if they have an activity, and lose those bonus moves when the activity is complete and their activity backlog is empty

Update do_turn_move_loot to support NPCs instead of the player performing the activity. Handle a weird NPC only corner case where the NPC is still active but not in the bounds of map by unloading them.

Change validate_sort_points to check if there is an unsorted loot zone, sorted loot zone, and camp food zone nearby. Change the Distribute Food pseudo mission to take food from the camp food zone. Change the Menial Labor mission to have the NPC perform the loot sorting activity using the infrastructure set up earlier in this PR.

Describe alternatives you've considered

Leaving the NPC in the overmap buffer and copy-pasting the do_activity_move_turn code into menial_return is easier but not the best solution.

Additional context

[mlangsdorf]

Current status: the NPC does some sorting, taking up time and freezing PC activity. I don't think that's right.

I think the issue is that activity_on_turn_move_loot is doing too much per cycle. I want it to do exactly 1 turn's worth of work per invocation, and I don't think it's doing that. I think it's sorting one entire tile and then getting invoked again.

[ifreund]

The problem is that move_item here modifies the player's moves not the npc's, so this check never succeeds

Also, letting npcs sort may cause some strange behavior if the player defines new zones during sorting. I'm pretty sure that because of the caching any new LOOT_UNSORTED zones will simply be ignored, and new destination zones will be properly sorted to if they are closest, but it would be good to investigate.

Edit: Looks like new unsorted zones will be sorted from but not make use of the caching. zone_manager::get_num_processed() could be modified to add any new zones to the cache.

[mlangsdorf]

@ifreund if you have a chance, please review or playtest. thanks!

[ifreund]

Looks good, still need to fetch and test though.

Changing zone_manager::get_num_processed() to

int zone_manager::get_num_processed( const tripoint &src )
{
    auto it = num_processed.find( src );
    if( it != num_processed.end() ) {
        return it->second;
    }
    // A zone was added while sorting, so add it to the cache.
    return num_processed[src] = 0;
}

should fix the minor caching issue I brought up.

Edit: Need to de-constify get_num_processed then of course.

[ifreund]

Shoot, somehow it took me until now to realize that this actually totally breaks caching if the player decides to remove items from the unsorted zone. What we need to do is clear the cache then restart the activity one extra time the first time we think we are done. There's a low chance that could still leave some items unsorted, but I think that's not a huge deal.

[mlangsdorf]

I will address your caching concerns tomorrow when I am slightly less brain-fried.

[ifreund]

Alright just to flesh out what I said about caching a bit:

• Adding items to a zone while sorting is no problem, as items are added to the top and sorting checks from bottom to top in order.
• Removing items that had been cached as unsortable will cause an equal number of sortable items on that tile to be skipped.
• The most complete fix is to always restart the activity using mgr.start_sort() and mgr.end_sort() to clear the cache unless we iterated through all the items and reached the end of the sort without restarting the activity. If that happens then all remaining items are unsortable and we have completed the activity.

Hopefully that all makes sense.

[mlangsdorf]

jenkins rebuild

[mlangsdorf]

Resolved the caching issue by calling end_sort() on any exit from the function.

Added logic to handle the NPC leaving the reality bubble by unloading them and then giving them a lot of bonus moves when they're loaded back into the reality bubble. It seems to work in my tests but could use more testing.