add_agent!() adds agents with out-of-bounds coordinates despite BoundsError

[Faroukbed]

Describe the bug
When add_agent!() is called with out-of-bounds coordinates on a GridSpace, it throws a BoundsError but still adds the agent to the model's dictionary. This creates an inconsistent state where the agent exists in the model but it has an invalid position. Functions such as nearby_ids() do not return this agent's id. I have not tested with other functions though.

Minimal Working Example

using Agents

@agent struct Customer(GridAgent{2}) 
    age
end

space = GridSpace((100,100); periodic = false, metric = :euclidean)
model = StandardABM(Customer, space)
add_agent!((0,0), model,25) #similar behavior when using negative coordinates

Output:

BoundsError: attempt to access 100×100 Matrix{Vector{Int64}} at index [0, 0]

However when checking the agents in the model the agent has been added

allagents(model)

Output:

ValueIterator for a Dict{Int64, Customer} with 1 entry. Values:
  Customer(1, (0, 0), 25)

Agents.jl version

Agents v6.2.10

[Ranakghosh7]

I'll start by writing a test case to prove that out-of-bounds agents currently get stuck in the model despite the error.

Then, I'll fix add_agent! by validating coordinates first, ensuring the function fails safely without corrupting the model state.

[Vyshnavi0702]

Hi! I would like to work on this issue if it is still open.

My plan is:

1. Add a test to reproduce the out-of-bounds agent behavior.
2. Validate coordinates before adding the agent to ensure the model state is not modified on error.

Please let me know if this approach is okay or if you suggest any changes.
Thank you!