Add support for or in ECS querying

[tristanpemble]

In investigating work for #63 it came up that we cannot do an or query right now. The specific query I wanted was:

• Has A, B, and C
• And at least one of those changed

It would be nice to be able to use some form of or statement when querying.

[Emily]

Summing up some conversation from the Discord on this topic:

• Does Any<(Component1, Component2)> make sense?
• Any seems most obviously valuable when dealing with Changed/Mutated/Added.
• Perhaps AnyChanged<(C1, C2)> (and mutated/added versions) would make more sense?
• It seems a little bit weird to implement AnyChanged when a more powerful primitive Any could be implemented.
• Tuple approach is nice on AnyChanged as it enables many components to be queried at once.
• How would the Any API look?
• Query<(C1, Option<C2>)> already exists

I'd like to take a stab at implementing the AnyChanged<(C1, C2, ...)> approach. It'd result in queries that look something like:

fn my_system(query: &mut Query<(Component1, AnyChanged<(Component2, Component3))> {
    for (component1, (component2, component3)) in &mut query.iter() {
        // do stuff

        if component2.changed {
            // do more stuff
        }
    }
}

I'm not sure if adding the changed flag to Changed is really all that helpful, but it's easy to do and shouldn't have much downside -- other than being a bit confusing in a context like Query<Changed<Component1>> where it'd always be set to true.

[Emily]

Some more Discord conversation summary:

• adding changed/mutated/added flags to Changed/Mutated/Added can cause some oddities around derefencing
• use a static fn to access the flag, e.g. if Changed::changed(c1) {
• also it doesn't really make sense to modify Changed and friends, just use new types
• naming new types is hard, but it makes more sense than something like let (c1_changed, c2_changed) = AnyChanged::changed(r);

[BimDav]

Sorry I did not see your recent comments, Emily. I tried a solution using directly the should_skip methods of two queries to implement Either<Query1, Query2>

[Emily]

Even though I really like the tuple approach I outlined, it has a ton of implementation complexity. @BimDav i think your Either implementation is the right way to go.

[cart]

Closed by #358