Stop the backend suites logging captured output (#22581)

Since #22246 moved these suites to OSDC, `arm_ethos_u55 / models` and
`arm_ethos_u85 / models` have died partway through reporting their
failures. The cause is the runner agent's heap, not the pod: the log
ends with `##[error]Exception of type 'System.OutOfMemoryException' was
thrown.` Each Ethos-U failure captures a few hundred thousand lines of
Vela operator listings, and the agent cannot process a step that size.

`--show-capture=no` drops those sections from the terminal report.
Verified on the default label and worker count:
https://github.com/pytorch/executorch/actions/runs/33935989761

```
success  arm_ethos_u85, models     <- never passed on OSDC before
success  arm_ethos_u55, models
9 failed, 13 passed, 20 skipped
log: 12,596 lines                  <- was 458k, truncated where the agent died
```

Nothing is lost. `--show-capture` only affects what the terminal prints,
so the captured streams still ship in the JSON report artifact (19MB,
against 17MB before), and each failure keeps its traceback and exception
message — `RuntimeError: Corstone simulation failed` and the arena sizes
are still in the log.

This applies to every backend suite, Linux and macOS, since
`_test_backend.yml` sources the script in both jobs. Only the Ethos-U
cells produce output at this scale today, but any of them could.

The failures themselves are pre-existing and unrelated; the job was
green with them before because `test_backend.sh` exits with the summary
generator's status rather than pytest's.

Authored with Claude Code.

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils
@Sebastian-Larsson @robell @rascani

Author: huydhn

.ci/scripts/test_backend.sh

@@ -158,7 +158,12 @@ GOLDEN_DIR="${ARTIFACT_DIR}/golden-artifacts"
 export GOLDEN_ARTIFACTS_DIR="${GOLDEN_DIR}"
 
 EXIT_CODE=0
-PYTEST_ARGS=(-c /dev/null -n auto)
+# An Ethos-U failure captures a few hundred thousand lines of Vela operator
+# listings, and the runner agent throws System.OutOfMemoryException processing
+# a step that size, taking the whole job down before pytest can report. The
+# reason for each failure is in its exception message and traceback, which are
+# unaffected.
+PYTEST_ARGS=(-c /dev/null -n auto --show-capture=no)
 if [[ ${#PYTEST_RETRY_ARGS[@]} -gt 0 ]]; then
     PYTEST_ARGS+=("${PYTEST_RETRY_ARGS[@]}")
 fi