SCP 3619: costing for serialiseData (#4480)

* Cost function for serialiseData, improved modelling method for fan-shaped data

* Minor corrections

* Tidying up

* Make info messages clearer

* Tidying up

* Tidying up

* Expand comment

* Typo

plutus-core/cost-model/budgeting-bench/Benchmarks/Data.hs

@@ -129,7 +129,10 @@ benchEqualsData =
 benchSerialiseData :: Benchmark
 benchSerialiseData =
     createOneTermBuiltinBench SerialiseData [] args
-        where args = dataSampleForEq -- FIXME: is this a good sample for serialization?
+        where args = dataSampleForEq
+    -- FIXME: see if we can find a better sample for this. More generally, how
+    -- does the internal structure of a Data object influence serialisation
+    -- time?  What causes a Data object to be quick or slow to serialise?
 
 makeBenchmarks :: StdGen -> [Benchmark]
 makeBenchmarks gen =

plutus-core/cost-model/data/builtinCostModel.json

@@ -198,8 +198,8 @@
     "equalsData": {
         "cpu": {
             "arguments": {
-                "intercept": 1083531,
-                "slope": 12717
+                "intercept": 1171539,
+                "slope": 18595
             },
             "type": "min_size"
         },
@@ -505,8 +505,8 @@
     "serialiseData": {
         "cpu": {
             "arguments": {
-                "intercept": 1000,
-                "slope": 255499
+                "intercept": 1194288,
+                "slope": 389135
             },
             "type": "linear_cost"
         },