test(zebrad): cap the live-sync tests at 2 minutes per attempt

[upbqdn]

Motivation

Closes #11399.

Solution

Lower the per-attempt slow-timeout for integration::sync::sync_one_checkpoint_mainnet
and integration::sync::restart_stop_at_height from 10 minutes to 2, keeping their two
retries, and record in .config/nextest.toml and on sync_until why nextest has to be
the bound: with MempoolBehavior::ShouldNotActivate, sync_until waits for zebrad to
exit with wait_with_output, which TestChild::with_timeout does not bound.

Tests

These two tests sync from live peers, so a slow peer stalls them with no limit of their
own. A run on the changed config reproduced a stall and shows the new bound working: the
attempt was terminated at 120s and the retry passed in 7.6s, for 138s total where the old
600s cap would have spent about 610s.

  TRY 1 SLOW [> 60.000s] zebrad::zebrad-tests integration::sync::sync_one_checkpoint_mainnet
TRY 1 TRMNTG [>120.000s] zebrad::zebrad-tests integration::sync::sync_one_checkpoint_mainnet
   TRY 1 TMT [ 120.003s] zebrad::zebrad-tests integration::sync::sync_one_checkpoint_mainnet
  TRY 2 PASS [   7.598s] (2/2) zebrad::zebrad-tests integration::sync::sync_one_checkpoint_mainnet
     Summary [ 138.188s] 2 tests run: 2 passed (1 flaky), 1334 skipped

Both tests pass unstalled in 7.6-13.7s, so 2 minutes leaves roughly nine times the
runtime they need. Making the wait itself bounded, so a stall fails where it happens
instead of relying on the test runner, is left to #11399's follow-up discussion.

AI Disclosure

• No AI tools were used in this PR
• AI tools were used: Claude, to measure the stall and write the change.

PR Checklist

• The PR title follows conventional commits format: type(scope): description
• The PR follows the contribution guidelines.
• This change was discussed in an issue or with the team beforehand.
• The solution is tested.
• The documentation and changelogs are up to date.

[v12-auditor]

Warning

No auditable source files found in this PR's diff.