GH-42173: [R][C++] Writing partitioned dataset on S3 fails if ListBucket is not allowed for the user (#47599)

### Rationale for this change

This PR gives the user to choose not to create directory in the bucket before writing dataset.
In case the `create_directory` option is set to FALSE, no verification will be made by R arrow.
The S3 storage will itself verify if the directory exists and if the users has the rigth to modify it.
This way no `ListBucket`  or ` HeadBucket`  are necessary to achieve the write operation. 

``` 
df |> arrow::write_dataset(
  minio$path(paste0("smartsla-bucket/rarrow/")),
  partitioning = "qualitative",
  create_directory = FALSE,
  format = "parquet"
)
```

### What changes are included in this PR?

`create_directory` is now available to the user in the `write_dataset` function.
Before this PR, this option was automatically set to TRUE (by default).

### Are these changes tested?

Yes

### Are there any user-facing changes?

No, the default value for `create_directory` is still TRUE.

* GitHub Issue: #42173

Lead-authored-by: Simon ELBAZ <[email protected]>
Co-authored-by: Simon Elbaz <[email protected]>
Co-authored-by: Antoine Pitrou <[email protected]>
Signed-off-by: Antoine Pitrou <[email protected]>

Author: simonelbaz

r/R/arrowExports.R

@@ -67,6 +67,10 @@
 #' group and when this number of rows is exceeded, it is split and the next set
 #' of rows is written to the next group. This value must be set such that it is
 #' greater than `min_rows_per_group`. Default is 1024 * 1024.
+#' @param create_directory whether to create the directories written into.
+#' Requires appropriate permissions on the storage backend. If set to FALSE,
+#' directories are assumed to be already present if writing on a classic
+#' hierarchical filesystem. Default is TRUE
 #' @param ... additional format-specific arguments. For available Parquet
 #' options, see [write_parquet()]. The available Feather options are:
 #' - `use_legacy_format` logical: write data formatted so that Arrow libraries
@@ -132,6 +136,7 @@ write_dataset <- function(dataset,
                           max_rows_per_file = 0L,
                           min_rows_per_group = 0L,
                           max_rows_per_group = bitwShiftL(1, 20),
+                          create_directory = TRUE,
                           ...) {
   format <- match.arg(format)
   if (format %in% c("feather", "ipc")) {
@@ -224,7 +229,7 @@ write_dataset <- function(dataset,
     partitioning, basename_template,
     existing_data_behavior, max_partitions,
     max_open_files, max_rows_per_file,
-    min_rows_per_group, max_rows_per_group
+    min_rows_per_group, max_rows_per_group, create_directory
   )
 }
 

r/R/dataset-write.R

@@ -318,7 +318,7 @@ void ExecPlan_Write(const std::shared_ptr<acero::ExecPlan>& plan,
                     arrow::dataset::ExistingDataBehavior existing_data_behavior,
                     int max_partitions, uint32_t max_open_files,
                     uint64_t max_rows_per_file, uint64_t min_rows_per_group,
-                    uint64_t max_rows_per_group) {
+                    uint64_t max_rows_per_group, bool create_directory) {
   arrow::dataset::internal::Initialize();
 
   // TODO(ARROW-16200): expose FileSystemDatasetWriteOptions in R
@@ -335,6 +335,7 @@ void ExecPlan_Write(const std::shared_ptr<acero::ExecPlan>& plan,
   opts.max_rows_per_file = max_rows_per_file;
   opts.min_rows_per_group = min_rows_per_group;
   opts.max_rows_per_group = max_rows_per_group;
+  opts.create_dir = create_directory;
 
   ds::WriteNodeOptions options(std::move(opts));
   options.custom_schema = std::move(schema);