docs: JUnit rows from testlist run carry the test title and run duration

DOCUMENTATION.md

@@ -1030,7 +1030,7 @@ testsprite testlist run <list-id> --wait --timeout 600 --output json  # poll eve
 testsprite testlist run <list-id> --wait --report junit --report-file ./results.xml
 ```
 
-**`testlist run`** dispatches each case through its project's configured environment and returns a pollable `runId` per case (the same id `test wait` / `runs` accept). Without `--wait` it exits 0 once every case is accepted (a fire-and-return, like `test run` without `--wait`). With `--wait` it polls every run to terminal and the exit code reflects the batch: **0** all passed, **1** any failed, **4** a partial `--case` miss — some requested ids are not members of the list (checked after the poll, so a genuine failure or timeout on the matched subset is reported first; the missed ids are warned to stderr at dispatch and carried on `notFound` in the JSON), **6** nothing new was dispatched because every targeted case conflicted — causes are named individually (already in flight, view-only mirror project, un-runnable environment, unknown id, dispatch error), and this exits 6 on the non-`--wait` path too, **7** any run timed out (against `--timeout`, default 600 s) or was rate-deferred (also on the non-`--wait` path — a partial dispatch never passes CI silently). A per-member poll error surfaces its real exit code through the shared `--wait` precedence table (see `test run`). `--case <test-id>` (repeatable) runs a subset; an empty match (no accepted, no conflicts) returns a `reason` and exits 0. `--report junit --report-file <path>` writes a JUnit sidecar after polling (suite name defaults to `testsprite:testlist:<listId>`; override with `--report-suite-name`; unlike `test run --all` / `test rerun` reports, `<testcase>` rows currently carry the test id as the name and no duration). Under `GITHUB_ACTIONS=true` (or `--gh-output` to force it locally) a `--wait` run also emits annotations (`::error::` for failed/timed-out members, `::warning::` for never-dispatched ones) + a job-summary table, and `--summary-file <path>` writes the reduced machine summary JSON (`{total, passed, failed, skipped, timedOut, runs[]}`); both require `--wait`, and non-dispatched members (conflicted, not-found) fold into the summary as `skipped` rows — visible, but never counted as `failed` — so a partial run cannot read as all-passed and the artifact cannot contradict the exit code.
+**`testlist run`** dispatches each case through its project's configured environment and returns a pollable `runId` per case (the same id `test wait` / `runs` accept). Without `--wait` it exits 0 once every case is accepted (a fire-and-return, like `test run` without `--wait`). With `--wait` it polls every run to terminal and the exit code reflects the batch: **0** all passed, **1** any failed, **4** a partial `--case` miss — some requested ids are not members of the list (checked after the poll, so a genuine failure or timeout on the matched subset is reported first; the missed ids are warned to stderr at dispatch and carried on `notFound` in the JSON), **6** nothing new was dispatched because every targeted case conflicted — causes are named individually (already in flight, view-only mirror project, un-runnable environment, unknown id, dispatch error), and this exits 6 on the non-`--wait` path too, **7** any run timed out (against `--timeout`, default 600 s) or was rate-deferred (also on the non-`--wait` path — a partial dispatch never passes CI silently). A per-member poll error surfaces its real exit code through the shared `--wait` precedence table (see `test run`). `--case <test-id>` (repeatable) runs a subset; an empty match (no accepted, no conflicts) returns a `reason` and exits 0. `--report junit --report-file <path>` writes a JUnit sidecar after polling (suite name defaults to `testsprite:testlist:<listId>`; override with `--report-suite-name`; each `<testcase>` is named by the test's title, falling back to the test id when there is no title; members that ran carry their run duration (`time="0"` when timing is unavailable, as for never-dispatched or timed-out members); every row keeps the test id as a `testId` attribute, and rows with a run also carry `runId`). Under `GITHUB_ACTIONS=true` (or `--gh-output` to force it locally) a `--wait` run also emits annotations (`::error::` for failed/timed-out members, `::warning::` for never-dispatched ones) + a job-summary table, and `--summary-file <path>` writes the reduced machine summary JSON (`{total, passed, failed, skipped, timedOut, runs[]}`); both require `--wait`, and non-dispatched members (conflicted, not-found) fold into the summary as `skipped` rows — visible, but never counted as `failed` — so a partial run cannot read as all-passed and the artifact cannot contradict the exit code.
 
 Every mutation (`create`/`update`/`delete`/`add`/`remove`) is refused with a billing-hold envelope on a **paused** team workspace (matching the portal's read-only wall); `delete` also removes the list's schedules and their triggers, so it requires `--confirm`. All commands accept the global `--output json|text` and `--dry-run` (offline shape-accurate sample).