lib: relax type hints on flatten function

The documented type signature of the flatten function is <list<dyn>> ->
<list<dyn>>, but the implementation was <list<typeV>> -> <list<typeV>>.
This has an impact on uses of flatten where the input is a nested set of
lists, the exact intended use-case. In this situation, the type checker
is told that the returned value is a list of lists..., resulting in
rejection of valid programs. There is a work-around that can be used,
but instead, just make the implementation match the documentation and
intention.

lib/collections.go

@@ -358,8 +358,8 @@ func (collectionsLib) CompileOptions() []cel.EnvOption {
 		cel.Function("flatten",
 			cel.MemberOverload(
 				"list_flatten",
-				[]*cel.Type{listV},
-				listV,
+				[]*cel.Type{listDyn},
+				listDyn,
 				cel.UnaryBinding(catch(flatten)),
 			),
 		),

testdata/flatten_types.txt

@@ -0,0 +1,87 @@
+mito -use collections -data state.json src.cel
+! stderr .
+cmp stdout want.txt
+
+-- state.json --
+{
+	"aye": {
+		"value": 1
+	},
+	"bee": {
+		"value": "two"
+	},
+	"one": [
+		{
+			"value": 1
+		},
+		{
+			"value": "two"
+		}
+	],
+	"two": [
+		{
+			"value": 3
+		},
+		{
+			"value": "four"
+		}
+	]
+}
+
+-- src.cel --
+[
+	"one",
+	"two",
+].map(k, state[k].map(e, {
+	k: e,
+})).flatten().map(d, d.with({
+	"aye": state.aye,
+	"bee": state.bee,
+}))
+-- want.txt --
+[
+	{
+		"aye": {
+			"value": 1
+		},
+		"bee": {
+			"value": "two"
+		},
+		"one": {
+			"value": 1
+		}
+	},
+	{
+		"aye": {
+			"value": 1
+		},
+		"bee": {
+			"value": "two"
+		},
+		"one": {
+			"value": "two"
+		}
+	},
+	{
+		"aye": {
+			"value": 1
+		},
+		"bee": {
+			"value": "two"
+		},
+		"two": {
+			"value": 3
+		}
+	},
+	{
+		"aye": {
+			"value": 1
+		},
+		"bee": {
+			"value": "two"
+		},
+		"two": {
+			"value": "four"
+		}
+	}
+]