Rework error handling on engine_forkchoiceUpdated failures

[shotes]

Beacon-kit currently asynchronously pushes forkchoice updates to the Execution Layer client over the Engine API via go s.sendPostBlockFCU(ctx, st, cBlk). When this forkchoice update fails, the beacon-kit node will simply continue on processing blocks and not updating the EL client on the new finalized block and the current head of the chain.

In my opinion, beacon-kit cannot simply move on with processing blocks if the forkchoiceUpdate is failing for any reason. At the same time, it is difficult to suggest returning error in FinalizeBlock(). Beacon-kit should introduce error handling that attempts to recover from recoverable errors and prevents the node from participating until the issue with the Engine API is resolved.

A potential path forward would be to start the FCU in a goroutine that:

• retries on ephemeral error, with an exponential backoff that will timeout eventually
• errors when there is an unrecoverable failure
• prevents the beacon-kit node from participating (not proposing, verifying proposals, or processing new blocks) until the forkchoice is updated on the EL.