Store task stdout and stderr in blobs with task_id in names

We were naming task stdout and stderr blobs without task_ids but
with invoke ids. If a function is running multiple times per
invoke this results in overwriting stdout and stderr blobs
by each task. Also depending on timings Server DB could
store wrong blob sizes (cause blobs could get overwritten
after DB record was saved).

This all gets fixed just by adding task_id into stdout, stderr
blob names. Max S3 object name is 1024 bytes long. Storing two
uuids (invoke id and task id) uses < 100 bytes of this space.
So should be fine.

No other places need to get updated because the DB is keyd using
task_ids, and all Server APIs also include task ids in urls.

Testing:

make build
cargo test

server/src/routes/internal_ingest.rs

@@ -140,11 +140,12 @@ pub async fn ingest_files_from_executor(
                     IndexifyAPIError::bad_request("task_result is required before diagnostics")
                 })?;
                 let file_name = format!(
-                    "{}.{}.{}.{}.{}",
+                    "{}.{}.{}.{}.{}.{}",
                     task_result.namespace,
                     task_result.compute_graph,
                     task_result.compute_fn,
                     task_result.invocation_id,
+                    task_result.task_id,
                     name,
                 );
                 let res = write_to_disk(state.clone().blob_storage, &mut field, &file_name).await?;