Improve .model type hint

[dimaqq]

It seems that most charms' integration tests assume that ops_test.model is truthy (6592 call sites):

async def setup_env(ops_test: OpsTest):
    await ops_test.model.set_config(...)

And a few ensure that it's truthy (46 call sites):

async def test_integrate(ops_test: OpsTest):
    assert ops_test.model is not None
    await asyncio.gather(
        ops_test.model.integrate("mimir:s3", "s3"),
        ...
    )

I didn't find a single test that's conditional on the model being or not being there. We're tracking ~160 charming repos in https://github.com/tonyandrewmeyer/charm-analysis

Currently, it's typed so:

    @property
    def model(self) -> Optional[Model]:
        """Represents the current model."""
        current_state = self.current_alias and self._models.get(self.current_alias)
        return current_state.model if current_state else None

I'm proposing to type it as Model and raise if the model is not available.

[addyess]

@dimaqq Do you think it constitutes a major rev on this package? It feels kinda sneaky to keep subtly changing the API. I get the reasoning to begin to add a ModelMissing error -- but it feels a bit icky to change the API.

[dimaqq]

I’m about 98% confident that it’s a minor or micro version bump.

I haven’t found a charm that relies .model being nullable.

If there is such a charm, that’s a different story.

[addyess]

@dimaqq if it's critical you can post a PR and @ca-scribner or myself would review.

if you can wait a bit -- maybe i can get to it in the next few weeks