Make the HFC aware of Peras rounds

[geo2a]

Fixes tweag/cardano-peras#65

Changes

• Changes in the HFC types:
  • EraParams now keeps track of an optional Peras round length.
  • Bound now keeps track of an optional Peras round number.
  • In the Serialise instances for EraParams and Bound, the encoders generate different CBOR depending on whether of not the Peras-relate components are present. The decoders act differently depending on the length of the CBOR list.
  • The EraSummary, while not modified directly, is now Peras-aware via EraParams and Bound:
    • in a valid summary, Peras round length must divide the epoch size. See the changes in EraSummarys haddock and in invariantSummary's implementation.
• Changes in the HFC time translation queries:

  • Two new top-level queries are exposed from Ouroboros.Consensus.HardFork.History.Qry:

    perasRoundNoToSlot :: PerasRoundNo -> Qry SlotNo
    slotToPerasRoundNo :: SlotNo -> Qry PerasRoundNo

  • add a roundtrip test that ensures that converting Peras round number to a slot and then back is an identity is added into the Test.Consensus.HardFork.Summary module.

  • add a Peras-specific test into Test.Consensus.HardFork.History module.

Suggested review order

Even though this PR affects many files, the bulk of the changes is localised. I suggest the following review order:

• First review the datatype and codec changes in ouroboros-consensus/Ouroboros/Consensus/HardFork/History/EraParams.hs and ouroboros-consensus/src/ouroboros-consensus/Ouroboros/Consensus/HardFork/History/Summary.hs
• Proceed to review the time translation query changes in ouroboros-consensus/src/ouroboros-consensus/Ouroboros/Consensus/HardFork/History/Qry.hs
• Then review the newly added tests in consensus-test/Test/Consensus/HardFork/History.hs and consensus-test/Test/Consensus/HardFork/Summary.hs

[amesgen]

Drive-by comment: Why does this need to show up here (and lots of other places)? We didn't have to do this when preserving compatibility for GenesisWindow.

[geo2a]

You are right, is is not needed here. I got entangled in typeclasses and I'm in the process of getting disentangled.

One difference from GenesisWindow is that PerasRoundLength unfortunately needs to be mentioned in era bounds (the Bound type), so their serialization needs to change as well. Hence, there will be bit a little more calls to given, but I'll work on minimizing their amount and removing the redundant ones.