fix(scheduler): Concurrent Bootstrap

[christ-tt]

Problem

• in _process_join, we check if we've received enough nodes (> min_nodes), and if so, we call bootstrap();
• in _wait_for_bootwtrap, we perform this once again.
• we may thus see concurrent bootstrap.
• Similarly, as we handled global_rebalance per node leave, we may see concurrent reboot.

In this PR, we

• simplify how we check bootstrapped or not;
• fix concurrent bootstrap by removing the attempt from wait event;
• fix concurrent re-boot by moving global_rebalance to _process_leave
• make our logs more informative.

[christ-tt]

Pipeline readiness check will be introduced in this PR