reserve: tolerate guest downtime for the whole reservation

tmt gives up on a guest that stays unreachable for TMT_REBOOT_TIMEOUT seconds, 10 minutes by default. That default is a poor fit for a reserved machine: the user owns the box and is free to reboot it into an offline upgrade, a leapp run or a slow bare metal boot, all of which routinely take longer. When they do, the reservation task fails with a guest reconnect timeout and the machine is released, possibly hours before the reservation was supposed to end.

Default TMT_REBOOT_TIMEOUT to the reservation duration, so the guest may be down for as long as the reservation lasts. An explicit --tmt-environment TMT_REBOOT_TIMEOUT=... still wins.

Only reserve sets the default. request --reserve runs the user's tests first, so a test stuck in a reboot still fails after 10 minutes and the timeout stays with the user. restart --reserve adds no timeout, but updates one the CLI set for the original reservation, so a restart with a new --duration doesn't keep the old tolerance.

Resolves: TFT-5168 Assisted-by: Claude Code Signed-off-by: Miroslav Vadkerti mvadkert@redhat.com