JobItem drops statusNotes from job responses (per-row detail lost)

[jacalata]

JobItem.from_response at models/job_item.py:224 parses <notes> elements:

notes = [note.text for note in element.findall(".//t:notes", namespaces=ns)] or None

But the actual REST API response wraps notes in a <statusNotes> container with typed <statusNote> children — none of that is captured:

<statusNotes>
  <statusNote type="classifier" value="value" text="note" />
</statusNotes>

This is documented as the Query Job response schema — the doc lists enumerated type values like CountOfUsersAddedToSite, CountOfUsersSkipped, CountOfUsersWithInsufficientLicenses, CountOfUsersInformationUpdated, etc.

Reproducer

Verified live against Tableau Server 2025.1. POST /sites/<id>/users/import returns a job; polling GET /sites/<id>/jobs/<job-id> after completion returns XML like:

<job id="..." type="UserImport" ... finishCode="1">
  <statusNotes>
    <statusNote type="line" value="0" />
    <statusNote type="errorCode" value="1" />
    <statusNote type="message" value="..." />
    <statusNote type="username" value="unknown" />
  </statusNotes>
</job>

TSC's job.notes for this response is [].

Impact

Callers can't tell what happened at the row level. For user-import jobs, that means no way to distinguish "5 added, 1 skipped" from a general "job finished with finishCode=1". Any caller wanting to emit per-row status output has to bypass TSC and parse the raw XML themselves.

Proposed fix

Add a new structured status_notes property on JobItem, populated from <t:statusNotes>/<t:statusNote> with all three attributes preserved:

status_notes = [
    {"type": note.get("type"), "value": note.get("value"), "text": note.get("text")}
    for note in element.findall(".//t:statusNotes/t:statusNote", namespaces=ns)
]

Leave the existing notes: list[str] untouched for backwards compatibility.

Discovered while

Planning tabcmd createsiteusers --nowait / --silent-progress work (tableau/tabcmd#35). The current tabcmd 2 command prints one line per row imported; switching to bulk_add + wait_for_job for Classic parity needs per-row status back — which requires this data.