accumulators: Set collation based on output DfType

For `group_concat()` and `json_object_agg()`, we were hardcoding the
output collation to UTF-8. It's easy enough to figure out the actual
intended collation, we just need to look at the `DfType` that's passed
in to the `over()` function. We still default to UTF-8 lacking
anything else.

Fixes: REA-5948
Release-Note-Core: Use the intended collation for string accumulating
  aggregation function.
Change-Id: Ia2486de5838c162d23cbbd57538faeff90812c0e
Reviewed-on: https://gerrit.readyset.name/c/readyset/+/10442
Reviewed-by: Michael Zink <michael.z@readyset.io>
Tested-by: Buildkite CI

Author: jasobrown-rs

readyset-dataflow/src/ops/grouped/accumulator.rs

@@ -39,17 +39,19 @@ impl Accumulator {
         src: NodeIndex,
         over: usize,
         group_by: &[usize],
-        _over_col_ty: &DfType,
+        over_col_ty: &DfType,
         _dialect: &Dialect,
     ) -> ReadySetResult<GroupedOperator<Accumulator>> {
+        let collation = over_col_ty.collation().unwrap_or(Collation::Utf8);
+
         let out_ty = match &op {
             AccumulationOp::GroupConcat { .. } => {
                 antithesis_sdk::assert_reachable!("Accumulation::GroupConcat");
-                DfType::Text(/* TODO */ Collation::Utf8)
+                DfType::Text(collation)
             }
             AccumulationOp::JsonObjectAgg { .. } => {
                 antithesis_sdk::assert_reachable!("Accumulation::JsonObjectAgg");
-                DfType::Text(Collation::Utf8)
+                DfType::Text(collation)
             }
         };