* Logging a warning when deleting the entity and it is not in any
Region.
* Searching every single region for an entity that is in none of them.
Also:
* Add workaround for bug in specs where deleting any entity clears the
components before checking that the generation is correct (surprised
that we haven't encounted bugs from this yet).
* Properly update `tracked_entities` inside `RegionMap` when deleting
an entity. Previously, we just relied on this being updated in
`RegionMap::tick` by the absence of the `Pos` component at that index.
However, this means if a new entity is created at that index after
deletion and before calling `RegionMap::tick`, then this can be
interpreted as an entity moving between regions rather than one being
deleted and another created. It seems like this could lead to
synchronization bugs like not creating the new entity on the client
(although I haven't noticed this before, I think maybe we use newly
inserted `Uid`s to detect new entities rather than the region system?).
I think it could at least lead to sending redundant messages to
synchronize the new entity.
because: The test_economy_moderate_standalone was failing tests on
arm64 MacOS systems with the following error:
```
thread 'site::economy::context::tests::test_economy_moderate_standalone' panicked at 'assertion failed: site.economy.pop >= env.targets[&id]', world/src/site/economy/context.rs:562:17
```
This was originally reported in Discord by @sharp in July, 2022
https://discord.com/channels/449602562165833758/797421546871324712/997332351195553874
Debugging showed that the grass target was 880, and the population was only around 718.
Local testing showed that lowering the grass target to 700 fixed the issue.
this commit: Lowers the grass target to 700 in the standalone economy tests.
unaddressed questions:
- Why is the grass target so high in the standalone economy tests?
- Alternately, why did the grass population drift so low, and no longer aligned with the tests.
- Why is this passing in CI, but not local testing?
- There are a number of commented out add_settlement calls in the standalone economy tests. Were there issues with the tests prior?
