Log started & finished events for Immunisation Import

[chrislo]

This adds logging for the point where an ImmunisationImport is created (typically in ImmunisationImportsController#create) and the point where its processing is completed.

We chose to associate the logging with the ImmunisationImport model rather than the ProcessImportJob, partly because we're only interested in ImmunisationImport at the moment (not the other import types) and partly because some of the other imports span multiple jobs. Ultimately we care about the end-to-end user journey from when they initiate an import to when the system is ready for them to proceed.

We're using SemanticLogger rather than directly talking to AWS Cloudwatch Logs, because it feels like it reduces coupling between the Rails app and other systems.

We're relying on Semantic Logger's built-in functionality for recording the name of the class that logged the message to associate each of the logged events with the ImmunisationImport class. We're also relying on the timestamp that is automatically recorded for every event. This should be accurate enough for our purposes.

We've chosen to calculate the duration (in milliseconds) and include it in the "finished" event rather than having to calculate it in Cloudwatch/Grafana from the difference between the timestamp on the "started" and "finished" events. We're also recording the number of rows so that we can "normalize" the durations of the imports.

Like in Metrics::ExportSchoolMovesCountJob, we're recording the workgroup of the Team associated with the ImmunisationImport.

For the moment we've explicitly decided not to add any specs for these changes, because we think the existing specs exercise the logging sufficiently to be confident it doesn't break anything. Our thinking it that we'll add specs later once we've got a Cloudwatch/Grafana dashboard to reference, i.e. to indicate that the app needs to log certain things in order to support the relevant dashboard.

We've tested the logging in development by wiring it up to a manually created Cloudwatch Log Stream via a Semantic Logger Cloudwatch Logs appender with the appender format set to :json and we see log events like this:

    {
      "host": "skua",
      "application": "Semantic Logger",
      "environment": "test",
      "timestamp": "2026-04-29T15:30:42.721497Z",
      "level": "info",
      "level_index": 2,
      "pid": 84031,
      "thread": "5584",
      "named_tags": {
        "id": 1,
        "team_workgroup": "w1"
      },
      "name": "ImmunisationImport",
      "message": "started"
    }

    {
      "host": "skua",
      "application": "Semantic Logger",
      "environment": "test",
      "timestamp": "2026-04-29T15:30:47.982389Z",
      "level": "info",
      "level_index": 2,
      "pid": 84031,
      "thread": "5584",
      "named_tags": {
        "id": 1,
        "team_workgroup": "w1"
      },
      "name": "ImmunisationImport",
      "message": "finished",
      "payload": {
        "duration_ms": 5259,
        "count": 11
      }
    }

Jira Issue - MAV-7301

[floehopper]

@thomasleese Since you approved this, we;ve made a small change to use ImmunisationImport#rows_count instead of calling #count on ImmunisationImport#rows. Hope that's OK! 🤞