E2E: nightly and tag jobs keep the 215-minute limit that cut off PR cleanup #5261

The nightly and tag E2E jobs can run out of time during sandbox cleanup, the same way a PR job did. #5260 raises the limit only for the PR and fork E2E jobs, from 215 to 265 minutes. Nightly and tag jobs still use 215.

In this PR run the tests passed, but teardown started at 211m15s. Successful cleanups on the same lane took between 8 and 11.5 minutes, so the job hit its limit before cleanup finished. I have not seen this failure on a nightly or tag run yet. Their full-suite path is the same, so a run that long would fail there too.

There is a second gap in the fork workflow. Its Collect report step has no timeout-minutes of its own, while the same step in pull-requests.yaml is capped at 30. A collector that hangs there can spend the rest of the job budget before teardown runs.

Neither change was in #5260 because it only covers the lane that failed. Both need a decision on how much headroom the nightly and tag jobs should get.