refactor: create AlarmCoordinator to separate HA functions from the fetching logic

[palazzem]

Related Issues

• Follow-up for Data loss (no updates) when the connection is unstable #51

Proposed Changes:

This change refactors entirely how the DataUpdateCoordinator is exposed to the integration. Instead of writing the fetching logic inside the async_setup_entry as a standalone function, we moved the entire logic in a dedicated class that is easy to test. AlarmCoordinator is fully tested and a regression test is added to the test suite.

Furthermore, the integration now uses async_config_entry_first_refresh() to call the first update, so that if there are issues with the update, the integration will raise ConfigEntryNotReady exception, preventing the integration from loading. At that stage, the integration is flagged as "Failed to load", preventing logs spamming.

Testing:

• Install the integration
• It should work as usual as this is just a refactoring

Extra Notes (optional):

n/a

Checklist

• Related issues and proposed changes are filled
• Tests are defining the correct and expected behavior
• Code is well-documented via docstrings

[palazzem]

Changes are required to improve the quality.