Problem description

right now we have AttestationPool of SingleAttestation and AggregatedAttestationPool of aggregated Attestation. When producing block, we only query AggregatedAttestationPool. We should also scan through AttestationPool so that:

• lodestar has more attestations in its produced block, that brings more consensus value produced
• it's easier to identify lodestar block

cc @nflaig

Solution description

• enhance AttestationPool so that it does not only preaggregate attestations but also store validated SingleAttestation instances
  • index by committeeValidatorIndex so that it's easier to query
• when producing block:
  • loop through slots
  • for each slot, loop through att_data of both pools. Make sure all att_data pass validation in order not to include conflicted attestations
  • for each att_data, loop through committee index
  • merge data from both pools in getAttestationsForBlockElectra

Additional context

Note:

• this will take more time, but we already produced attestations while waiting for responses from EL/builder in feat: minimal changes to parallelize block building steps #7814
• Teku time boxed the process, maybe we don't need it as I'll suppose it does not take so much time as we should loop through AggregatedAttestationPool first and left only few "not seen attesters" but could be a nice too have when network is unstable https://github.com/Consensys/teku/pull/9487/files#diff-b0952debf32597dcee62e41a8381d3d1ca354b519b3eec1995be0651e7a0f616R516
  • we can bound it by some constants as we already did in AggregatedAttestationPool
• Analyze memory consumed. Each SingleAttestation = 240 bytes but it's way more than that due to NodeJS inefficiency to store Uint8Array (this happen to the signature). Good news is AttestationData is shared in the SeenAttestationDatas cache