Follower replica leader oversight

It's possible for players to leave behind a follower entity enough for entity sleep/networking range, while leaving said entity within range of another player, meaning they won't teleport.

That in itself is another deal and is technically intentional, however, the issue in question is that non-leader players won't know who the follower entity's leader is, due to entity sleep/networking range despawning entities locally for clients.

This results in being able to accidentally attack them when force-attacking with keybinds, which normally shouldn't happen otherwise. This also applies with non-player leaders and so, it's not just restricted to players (but normally only the player tag on the leader is checked, so technically nothing should be affected in the same way).

Obviously, this won't lead to Winona's catapults or things like that to accidentally hit them, since that's server-sided and the server has all the necessary data, this is specifically for clients selecting a target and syncing it to the server, and similar.

It's a very specific case, but it's relevant enough I feel due to potential accidents it could lead to.

This could be simply solved by doing something as shown below, and changing some of the checks in combat_replica and other similar places a bunch to account for the changes (search for replica.follower, there aren't that many cases of it).

local Follower = Class(function(self, inst)
    self.inst = inst

    self._leader = net_entity(inst.GUID, "follower._leader")
    self._leaderisplayer = net_bool(inst.GUID, "follower._leaderisplayer")
end)

function Follower:SetLeader(leader)
    self._leader:set(leader)

    local isplayer = leader ~= nil and leader:IsValid() and (leader.isplayer or leader:HasTag("player"))
    self._leaderisplayer:set(isplayer and true or false)
end

function Follower:GetLeaderIsPlayer()
    return self._leaderisplayer:value()
end

Of course, another simple alternative would be to set a unique tag on the follower and check for that instead, but that there's a limit to tags, this might not be ideal. These are just suggestions.

If other tags were to be used with follower replica leader checks, you could instead sync some specific tags with a net_array, but that's more expensive and it might not be needed, plus you'd need to account for the tags being added or removed.