UTxO-HD heap increase in latest benchmarks

[jasagredo]

Live data increased from 3800MB to 4200MB in value workload

and from 3400MB to 4600MB in plutus workload

[jasagredo]

Local cluster runs seem unreliable to reproduce this behavior. As the difference is more drastic in the forge-stress-pre-plutus workload we are running the following clusters:

• utxo-hd-9.1 with -hT
  
• utxo-hd-9.1 with -hi and -finfo-table-map -fdistinct-constructor-tables Nix needs adaptation to build with info table profiling.
• 9.1.0 with -hT
  

It is important to note that the max memory usage seems to be the Genesis loading phase, which influences the heap size for the overall run:

[jasagredo]

As a side note, #1194 doesn't seem to have any notable benefit in the memory usage.

[jasagredo]

The graphs above show that on local runs, there is only an 80 MB difference on BabbageTxOut which cannot be unpacked in GHC 8.10.7, see

ouroboros-consensus/ouroboros-consensus-cardano/src/ouroboros-consensus-cardano/Ouroboros/Consensus/Cardano/Ledger.hs

Line 94 in 4408285

| BabbageTxOut !(Value (LedgerState (ShelleyBlock (Praos c) (BabbageEra c))))

[jasagredo]

This is blocked on us being able to get profiling information from the cluster runs. We need the help from the Perf team.

[jasagredo]

Done in #1232