scripts: csv.py: Ignore missing by fields in enumerate/accumulate

Note this matches the behavior of mods, e.g. I would expect this to not
break if ORDER is missing:

  ./scripts/csv.py \
      -bcase='%(case)s+%(probe)s+%(ORDER)s' \
      -ft=accumulate(bench_simtime, case, probe, ORDER)

Normally the expr compiler would force typechecking of ORDER, giving it
a default value of int(0) if missing, but we intentionally bypass
typechecking in enumerate/accumulate's by fields since they may be
strings.

scripts/csv.py

@@ -812,12 +812,8 @@ class CsvExpr:
                 return CsvInt(0)
 
             # enumerate
-            k = ['enumerate', id(self)]
+            k = ('enumerate', id(self)) + tuple(
-            for v in self:
+                    fields.get(v.a) for v in self)
-                if v.a not in fields:
-                    raise CsvExpr.Error("unknown field? %s" % v.a)
-                k.append(fields[v.a])
-            k = tuple(k)
             x = state.get(k)
             if x is None:
                 x = 0
@@ -852,12 +848,8 @@ class CsvExpr:
                 return y
 
             # accumulate
-            k = ['accumulate', id(self)]
+            k = ('accumulate', id(self)) + tuple(
-            for v in it.islice(self, 1, None):
+                    fields.get(v.a) for v in it.islice(self, 1, None))
-                if v.a not in fields:
-                    raise CsvExpr.Error("unknown field? %s" % v.a)
-                k.append(fields[v.a])
-            k = tuple(k)
             x = state.get(k)
             if x is None:
                 x = y