Raise TMT_CONNECT_TIMEOUT to 1200s for metal boot times

tmt waits TMT_CONNECT_TIMEOUT seconds for a connection after guest boot. The default (tmt.guest.DEFAULT_CONNECT_TIMEOUT) is 120s and Testing Farm does not override it, so every worker runs with 120s.

Metal guests need far longer. Boot data from TFT-5015 puts public ranch generation 5/6 flavors at 706-820s and generation 7/8 at 202-583s. A metal guest that reboots during the run, or is still coming up when tmt starts waiting, loses the connection race and the run fails.

Same root cause as TFT-5015 in a different code path: TFT-5015 raised Ansible's reboot_timeout, this raises tmt's own post-boot connect wait.

Inject TMT_CONNECT_TIMEOUT=1200 via environment-variables in the test-schedule-tmt, test-schedule-tmt-connect and test-schedule-tmt-multihost configs, matching the 1200s TFT-5015 settled on. The value is a deadline, not a sleep, so a guest that never returns still fails as soon as tmt gives up retrying.

Applied to every ranch that provisions metal, including worker-redhat: it targets the osci artemis, which serves AWS .metal flavors and Beaker physical hardware (pools defined in ansible-baseos-ci/artemis-pools). Unlike TFT-5015's reboot_timeout, this is not public-ranch-only.

Related: TFT-5015, !1526 (merged)

Co-Authored-By: Claude Opus 4.8