Delete withdrawal bundle event block from the correct db on disconnect

[1440000bytes]

In disconnect, the withdrawal-event branch reads the sequence index from withdrawal_bundle_event_blocks.last() but deletes that index from deposit_blocks:

let (last_withdrawal_bundle_event_block_seq_idx, ...) = state
    .withdrawal_bundle_event_blocks
    .last(rwtxn)? ... ;
...
if !state
    .deposit_blocks
    .delete(rwtxn, &last_withdrawal_bundle_event_block_seq_idx)?
{
    return Err(Error::NoWithdrawalBundleEventBlock);
};

So on a reorg disconnecting a two-way peg block that carries a withdrawal bundle event, the event-block record is never removed, and an unrelated deposit record sharing the same sequence index is deleted instead (or the disconnect errors if no such deposit record exists). Both get_last_deposit_block_hash and get_last_withdrawal_bundle_event_block_hash drive mainchain sync position, so this corrupts deposit/withdrawal tracking and can diverge state across nodes.

Fix

Delete from withdrawal_bundle_event_blocks, matching the lookup and mirroring the deposit branch just below.

Test

Added a unit test that seeds a withdrawal_bundle_event_blocks row and a deposit_blocks row at the same index, disconnects a withdrawal bundle event and asserts the event-block row is removed while the deposit row survives. It fails on the old code and passes with the fix.