Let a jsonl export keep a record a value

`--format jsonl` existed, but `_cell` ran json.dumps at row-build time,
before a format was chosen — so a nested value was a string by then and
jsonl only re-escaped it, leaving a consumer against csv's 128KB field
limit either way. Stringifying moved to the csv writer, so csv is
byte-identical and jsonl carries json. The runs TUI followed, or a record
would draw as a Python repr.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KYM38KSb4V4v2T71eifnZv

backend/fluksio/api/routes/runs.py

@@ -406,12 +406,10 @@ def _selected(
 
 
 def _cell(value: Any) -> Any:
-    """A value as a table holds it: a scalar, or JSON when it is not one."""
-    if hasattr(value, "isoformat"):
-        return value.isoformat()
-    if value is None or isinstance(value, (str, int, float, bool)):
-        return value
-    return json.dumps(value)
+    """A value as a table holds it: a scalar, a date as its ISO string, or a
+    record kept as one — csv stringifies it at write time; jsonl wants it a
+    value, not a string full of it."""
+    return value.isoformat() if hasattr(value, "isoformat") else value
 
 
 def _leaves(value: Any, prefix: str = "") -> Iterator[tuple[str, Any]]:
@@ -464,13 +462,25 @@ def _drain(buffer: io.StringIO) -> str:
 
 
 def _csv(columns: list[str], chunks: Iterator[list[dict[str, Any]]]) -> Iterator[str]:
-    """Header first, then a chunk at a time through one reused buffer."""
+    """Header first, then a chunk at a time through one reused buffer.
+
+    A cell that is still a record here — `_cell` leaves one alone for jsonl's
+    sake — becomes the JSON string a csv cell can hold.
+    """
     buffer = io.StringIO()
     writer = csv.DictWriter(buffer, fieldnames=columns)
     writer.writeheader()
     yield _drain(buffer)
     for chunk in chunks:
-        writer.writerows(chunk)
+        writer.writerows(
+            {
+                key: value
+                if value is None or isinstance(value, (str, int, float, bool))
+                else json.dumps(value)
+                for key, value in row.items()
+            }
+            for row in chunk
+        )
         yield _drain(buffer)