Validator activation queue off by one

[clintjeff2]

Motivation

• Fix an off-by-one in validator activation scheduling where a pending validator with activation_epoch == current_epoch could be skipped and never activated.

Description

• Add a new deterministic activation queue in src/validator/activation_queue.rs with MAX_PENDING_VALIDATORS, compute_activation_epoch(), capacity/duplicate checks and an inclusive drain (activation_epoch <= current_epoch).
• Wire the activation queue into the validator API by exposing the module in src/validator/mod.rs and extending ValidatorSet in src/validator/validator_set.rs with a Pending status, an activation_queue field, add_pending_validator(), activate_validator(), queued_activations() and process_activation_queue() that activates drained validators exactly once.
• Add regression tests in tests/activation_queue_test.rs that verify boundary-epoch activation, multiple validators sharing the boundary, future delayed activation using compute_activation_epoch(), and duplicate/capacity rejection.

Testing

• Ran cargo test --test activation_queue_test and all activation-queue tests passed.
• Ran cargo test --test exit_queue_ordering_test and the existing exit-queue ordering tests passed.
• Ran the full test suite with cargo test and the repository tests completed successfully with no failures.

---

Closes #16