Better Typing support for (Cell)agents

[Corvince]

I recently enabled Pyright for my editor, its a python type checker which is much faster then mypy and therefore actually usable. At least thats my experience.

Anyway, my experience with Mesa was very humbling. Take for example this code snippet from the schelling benchmark model

        neighborhood = self.cell.neighborhood(radius=self.radius)
        for neighbor in neighborhood.agents:
            if neighbor.type == self.type:
                similar += 1
            [...]
            self.model.happy += 1

Lets start from bottom to top. The first typechecking error comes from self.model.happy. Since we hardcoded the model to be of type mesa.Model, it doesn't know that model.happy exists and is an integer. This was rather easy to fix with my first commit, where the Agent/CellAgent now can take a generic parameter for the model class. So you just have to define your Agent as

class SchellingAgent(CellAgent["Schelling"])

The next issue was that the ´neighbor.type` was unknown. This was much harder to fix and occupied me for quite a while. There are now two hoops you have to jump through to make it work (if you are determined to enable type checking).
First you have to again add this information to the CellAgent as before

class SchellingAgent(CellAgent["Schelling", "SchellingAgent"])

and you have to define the cell and agent class when you define the OrthogonalGrid like so

        self.grid = OrthogonalMooreGrid(
            [height, width],
            torus=True,
            capacity=1,
            random=self.random,
            cell_klass=Cell[SchellingAgent],
            agent_class=SchellingAgent,
        )

I think this is not really optimal, but so far the only solution I could find. Maybe someone who knows the Python type system better then me can help improve this PR

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🔴 +24.0% [+23.4%, +24.4%]	🔵 -0.3% [-0.7%, +0.1%]
Schelling	large	🔴 +23.8% [+22.4%, +25.5%]	🔵 -4.7% [-6.4%, -2.9%]
WolfSheep	small	🔵 +1.3% [-0.5%, +3.1%]	🔵 +1.5% [+1.0%, +1.9%]
WolfSheep	large	🔵 +2.9% [+1.5%, +4.3%]	🔵 -0.4% [-2.5%, +1.6%]
BoidFlockers	small	🔵 +1.1% [+0.3%, +2.0%]	🔵 +2.7% [+2.0%, +3.4%]
BoidFlockers	large	🔵 +0.8% [-0.0%, +1.5%]	🔵 +1.6% [+1.0%, +2.1%]

[EwoutH]

Sorry for the late reply. In general, I like the idea of typing in major Mesa components. I don't know if I would like it to be included in user-facing examples (not this PR though). More importantly, I have to read up on typing in Python, especially Generic Types, TypeVar and the square brackets stuff.

I will try to do that in the upcoming week. How quick do you want to move with this? And what's your vision on typing in Mesa in general?

[quaquel]

I guess this PR is largely superseded by the recent work on grid spaces, and so can be closed. @Corvince let me know if there is still something here that should be kept. Otherwise, I'll close this somewhere over the course of next week.

[quaquel]

we have made substantial progress on typing so this seems largely fixed by now.