Benchmark against ledger codebase

With all the recent performance improvements to `test all` we unfortunately don't see an across-the-board improvement in ledger:
```
new constrained-generators
- cabal test all: 700 seconds
- STS property tests:
cardano-core
STS property tests
GOV tests
prop_GOVCERT (9074ms)
+++ OK, passed 100 tests.
prop_POOL (4162ms)
+++ OK, passed 100 tests.
prop_DELEG (6219ms)
+++ OK, passed 100 tests.
prop_ENACT (4080ms)
+++ OK, passed 100 tests.
prop_RATIFY (12896ms)
+++ OK, passed 100 tests.
prop_CERT (16671ms)
+++ OK, passed 100 tests.
prop_GOV (71468ms)
+++ OK, passed 100 tests.
prop_EPOCH (79967ms)
+++ OK, passed 100 tests.
Finished in 204.5444 seconds
master
- cabal test all: 640 seconds
- STS property tests:
cardano-core
STS property tests
GOV tests
prop_GOVCERT (9123ms)
+++ OK, passed 100 tests.
prop_POOL (4764ms)
+++ OK, passed 100 tests.
prop_DELEG (7134ms)
+++ OK, passed 100 tests.
prop_ENACT (4412ms)
+++ OK, passed 100 tests.
prop_RATIFY (15777ms)
+++ OK, passed 100 tests.
prop_CERT (19451ms)
+++ OK, passed 100 tests.
prop_GOV (52386ms)
+++ OK, passed 100 tests.
prop_EPOCH (83900ms)
+++ OK, passed 100 tests.
Finished in 196.9526 seconds
```
We'd like to take what we have in the ledger code-base and benchmark against that - or variants of it that behave similarly. In general setting up a proper benchmarking suite for this library is a good idea. See #5.