chore: make event_name mandatory in Framework

[dimaqq]

I'm sure that the event name is always set (except some units tests)
Meanwhile we've got this signature: Framework.__init__(..., event_name: str|None, ...)
And therefore self._event_name is optional too.

This PR makes this mandatory.

Open question remaining: whether ops-scenario / ops version needs to be bumped to keep the two in sync.
That would be necessary if ops-scenario ships with harness code and ops is not.
(I don't recall if ops / ops['testing'] has reached that stage)

[benhoyt]

What's the motivation for this change? Unless we have a very compelling reason, I don't think we should do it, as it's a breaking change and will actually break some charm tests (GitHub Actions exporter, Prometheus K8s, Sunbeam charms, and Traefik Route):

~/w/charm-analysis/.cache$ rg 'Framework\(' --sort=path -A5
github-actions-exporter-operator/tests/unit/test_charm.py
40:        self.harness._framework = ops.framework.Framework(
41-            self.harness._storage, self.harness._charm_dir, self.harness._meta, self.harness._model
42-        )
43-        self.harness._charm = None
44-        self.harness.update_config({"github_webhook_token": "foo"})
45-        self.harness.enable_hooks()
--
68:        self.harness._framework = ops.framework.Framework(
69-            self.harness._storage, self.harness._charm_dir, self.harness._meta, self.harness._model
70-        )
71-        self.harness._charm = None
72-        self.harness.update_config({"github_webhook_token": ""})
73-        self.harness.enable_hooks()
--
118:        self.harness._framework = ops.framework.Framework(
119-            self.harness._storage, self.harness._charm_dir, self.harness._meta, self.harness._model
120-        )
121-        self.harness._charm = None
122-        new_webhook_token = token_hex(16)
123-        self.harness.update_config({"github_webhook_token": new_webhook_token})

prometheus-k8s-operator/tests/unit/test_remote_write.py
292:            self.harness._framework = framework.Framework(
293-                self.harness._storage,
294-                self.harness._charm_dir,
295-                self.harness._meta,
296-                self.harness._model,
297-            )

sunbeam-charms/ops-sunbeam/ops_sunbeam/test_utils.py
822:    harness._framework = framework.Framework(
823-        ops.storage.SQLiteStorage(":memory:"),
824-        harness._charm_dir,
825-        harness._meta,
826-        harness._model,
827-    )

traefik-route-k8s-operator/tests/unit/conftest.py
79:    harness._framework = framework.Framework(
80-        harness._storage, harness._charm_dir, harness._meta, harness._model
81-    )
82-
83-    harness._charm = None
84-    harness.begin()

[dimaqq]

Oh, I didn't realise that charm tests instantiated the ops.Framework directly, I thought that surely they used the fake framework from Harness or something like that.

Given this, I guess the proposed change is a no-go.

Maybe let's discuss if we can get rid of this nit somehow... (docs? long deprecation period? a default like event_name="he-be-dragons", smth else?).

I consider it a nit because a usable Framework, that is a framework that can call the user handlers for events, must have an event associated with it, or an event set to call handlers.