Client-prediction tracking ("Client-mapped entities")

[RJ]

This is working fine, in the test and in my game (which has a misprediction cleanup system similar to the new docs describe).

I'm currently sending Entity::PLACEHOLDER where there is no prediction, so still need to decide on the packet format.
EDIT: pushed a change to encode compactly now

With a bit more code complexity, like a shared config between client and server, we can make it totally opt-in and zero cost if the feature isn't used, otherwise I think i can make it cost 1 bit per entity to mark if there's a predicted entity too.

This is to fix #67

[RJ]

i haven't added any sort of special way for the client to tell the server about predictions, imo that should be part of the game's existing custom protocol. all the game server needs to do is detect that, and update the prediction tracker.

[RJ]

If we made a new shared config struct that the client and server plugins take, this kind of feature can be opt in depending on the style of game. (perhaps we init_resource::<RepliconConfiguration>() and let consumers replace it if they dont like the defaults, so no other API changes for now)

how about something like:

#[derive(Default)]
struct RepliconConfiguration {
    // Makes the `PredictionTracker` available on the server, 
    // and supports client-predicted entity mapping.
    pub client_predictions: bool,
}

Another option might be to not insert the PredictionTracker resource by default, and make the user insert the resource on both client and server if they want to enable the feature (even though the client doesn't need it)

[codecov-commenter]

Codecov Report

Attention: 7 lines in your changes are missing coverage. Please review.

Comparison is base (4dbe6f1) 96.16% compared to head (3fedfb4) 95.53%.
Report is 1 commits behind head on master.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #78      +/-   ##
==========================================
- Coverage   96.16%   95.53%   -0.63%     
==========================================
  Files          13       13              
  Lines         834      874      +40     
==========================================
+ Hits          802      835      +33     
- Misses         32       39       +7     

Files	Coverage Δ
src/lib.rs	100.00% <ø> (ø)
src/server/replication_buffer.rs	99.12% <100.00%> (+0.07%)	⬆️
src/server.rs	95.92% <80.00%> (-1.18%)	⬇️
src/client.rs	88.27% <78.94%> (-1.57%)	⬇️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[RJ]

reverted the entity encoding commit and added a currently-failing test showing the edge case @Shatur pointed out relating to mapped components referencing the predicted entity.

"include additional array with entity to entity mappings instead. This array should come first and will be processed before replication and it just updates mappings on client. So additional callback won't be required. "

also noted there's no need for the prediction hit callback, since successful prediction means the predicted entity will receive (at least) the Replication component, so hits can be detected with a Query<Entity, <With<Prediction>, Added<Replication>>>

i will change it to using an entity mapping array as discussed.

[UkoeHB]

For some reason server.rs contains duplicate code from client_entity_map.rs. I am not going to transfer all my comments over to the correct file.