Validator client shouldn't request aggregate with no aggregators

[michaelsproul]

Description

The validator client's logic for forming aggregate attestations will currently request an aggregate from the BN before checking whether there are any aggregators that need to sign it:

lighthouse/validator_client/src/attestation_service.rs

Lines 494 to 518 in 2841f60

	let aggregated_attestation = &self
	.beacon_nodes
	.first_success(
	RequireSynced::No,
	OfflineOnFailure::Yes,
	|beacon_node| async move {
	let _timer = metrics::start_timer_vec(
	&metrics::ATTESTATION_SERVICE_TIMES,
	&[metrics::AGGREGATES_HTTP_GET],
	);
	beacon_node
	.get_validator_aggregate_attestation(
	attestation_data.slot,
	attestation_data.tree_hash_root(),
	)
	.await
	.map_err(|e| {
	format!("Failed to produce an aggregate attestation: {:?}", e)
	})?
	.ok_or_else(|| format!("No aggregate available for {:?}", attestation_data))
	.map(|result| result.data)
	},
	)
	.await
	.map_err(|e| e.to_string())?;

This is inefficient, but also liable to cause issues with other clients, which may not be expecting to produce an aggregate when they are not subscribed to the attestation subnet. Thanks to @nflaig from Lodestar for finding and reporting this bug.

Steps to resolve

Check whether there are any aggregators (at least one selection proof that is Some) before requesting an aggregate from the BN.

[michaelsproul]

This also affects Lighthouse VC with a Prysm BN.

For easy searchability, the error produced is:

CRIT Error during attestation routine

[jimmygchen]

Completed in #4774.