Merge #2889 #2891

2889: Fix QueryAnytimeMary r=dcoutts a=Anviking

Before this fix, QueryAnytimeMary had the same implementation as
QueryAnytimeAllegra. It should presumably have an extra `S`.

### Context
@piotr-iohk noticed that the wallet's reporting of Mary-hard-fork-dates were wrong.

Running GetEraStart for all eras on mainnet we get
```
Byron: Right (Just (Bound {boundTime = RelativeTime 0s, boundSlot = SlotNo 0, boundEpoch = EpochNo 0}))
Shelley: Right (Just (Bound {boundTime = RelativeTime 89856000s, boundSlot = SlotNo 4492800, boundEpoch = EpochNo 208}))
Allegra: Right (Just (Bound {boundTime = RelativeTime 101952000s, boundSlot = SlotNo 16588800, boundEpoch = EpochNo 236}))
Mary: Right (Just (Bound {boundTime = RelativeTime 101952000s, boundSlot = SlotNo 16588800, boundEpoch = EpochNo 236}))
```
which is weird as the Mary fork here is the same as Allegra, instead of Nothing.

Haven't built or tested this. CI should tell though.

2891: Update to latest cardano-ledger-spec r=dcoutts a=nc6

This is something of a slapdash patch. Crucially, the latest changes in
the ledger are to support the generalisation of the update system for
privilege, and this means that the state of the update system will
change. At the moment, consensus drills deep into the update state to
work out when to execute a hard fork.

This PR locks down the update state to the Shelley update state. The
correct solution will be to work out exactly what consensus needs to
know and add an additional class to the API which provides this.

Replaces #2888

Co-authored-by: Johannes Lund <[email protected]>
Co-authored-by: Nicholas Clarke <[email protected]>

Author: 3 people