Add a test for relaxed ledger cost model serialization

[zliu41]

In ledger eras prior to Conway, we are unable to do either of these:

• Add new builtin functions to existing Plutus language versions (even with a HF)
• Add a new Plutus language version, unless we either put the new cost model in the genesis file (as we did for Plutus V1, and potentially also Plutus V3), or delay enabling the new Plutus version until one epoch after the HF (as we did for Plutus V2, but this will be more difficult to do going forward since the voting process may take much longer than one epoch).

For more details of the problem and the solution, see IntersectMBO/cardano-ledger#2902.

The solution described in the above issue has been implemented in ledger. We should have an end-to-end test that verifies that it indeed works.

cc @lehins @mkoura

[zliu41]

What needs to be tested: we need a test verifying that the following two versions of the node have the same behavior throughout PV9:

• The current version, where PlutusV2 has 175 cost model parameters
• The version to be released after merging this Plutus PR and this ledger PR, and adding one or more PlutusV3 builtins to PlutusV2, such that PlutusV2 has more than 175 cost model parameters (say 180)

where "the same behavior" means

• any PlutusV2 script not using the new primitives should work, and any PlutusV2 script using the new primitives should fail at phase 1, for both versions of the node

and "throughout PV9" includes

• At the beginning of PV9, where PlutusV2 cost model has 175 parameters
• After some PlutusV2 cost model parameters are updated, but the number of parameters remains 175
• After the number of PlutusV2 cost model parameters is updated to 180

@lehins @mkoura does this make sense?

[mkoura]

The tests are added and ran nightly.