docs: add descriptions for max_result_set_writers

Author: YoshiakiNishimura

docs/config-parameters.md

@@ -37,13 +37,14 @@ Target component
 | Parameter name | Type | Value | Remarks |
 |---:| :---: | :--- |---|
 | `commit_response` | String | Default value for commit waiting. Select one of the following: (`ACCEPTED`, `AVAILABLE`, `STORED`, `PROPAGATED`). The default is `STORED`. These symbols indicate the point in time until when the server delays notifying client of commit result. <ul><li>`ACCEPTED`: commit operation has been accepted, and then the transaction will never abort except system errors. </li><li>`AVAILABLE`: commit data has been visible for others. </li><li>`STORED`: commit data has been saved on the local disk. </li><li>`PROPAGATED`: commit data has been propagated to the all suitable nodes.</ul> | By explicitly specifying it at commit from the client, the above setting can be overwritten for each transaction. |
-| `default_partitions` | Integer | Number of partitions when data is divided for parallelizable relational operators. The default value is 5. |
+| `default_partitions` | Integer | Number of partitions when data is divided for parallelizable relational operators. The default value is 5. | The value must be equal to or less than `max_result_set_writers`. |
 | `lowercase_regular_identifiers` | Boolean (true/false) | Whether to convert table and column names into lowercase internally. The default value is false.||
 | `stealing_enabled` | Boolean (true/false) | Whether the scheduler steals tasks to utilize idle CPU cores. The default value is true. |
 | `thread_pool_size` | Integer | Number of threads used by the task scheduler in the SQL service. The default value is set according to the environment by the following formula. <ul><li>`MIN(<default worker coefficient> * <number of physical cores>, <maximum default worker count>)` </li><li>If the result is less than 1, it is set to 1. </li><li>the `<default worker coefficient>` is 0.8 </li><li>the `<maximum default worker count>` is 32.</ul> |
 | `scan_block_size` | Integer | The maximum number of scan operator records processed before yielding to other tasks. The default limit is 100 records. If set to 0, this limit is removed. The scan operator processes either the entire table or a specific range of data. This parameter is intended to prevent the thread from becoming unresponsive when the number of scan operator records is too large. |
 | `scan_yield_interval` | Integer | The maximum millisecond time of scan operator processing records before yielding to other tasks. The default is 1. If the value of this option is set to 0, the decision to split will rely solely on the value of scan_block_size. When using this option, it is recommended to keep the scan_block_size at its default value. If the scan_block_size is set to an extremely large value, the actual split occur after a time that significantly exceeds the specified value. |
-| `scan_default_parallel` | Integer | The maximum degree of parallelism for RTX scan tasks. The default value is 4. | This setting is experimental and may undergo changes in future releases. While it is experimental, this parameter must be enabled by also setting `dev_rtx_parallel_scan=true`. This parameter must be less than the value of default_partitions.|
+| `scan_default_parallel` | Integer | The maximum degree of parallelism for RTX scan tasks. The default value is 4. | This setting is experimental and may undergo changes in future releases. While it is experimental, this parameter must be enabled by also setting `dev_rtx_parallel_scan=true`. The value must be equal to or less than `max_result_set_writers`. |
+| `max_result_set_writers` | Integer | The maximum number of writers for a result set. The default value is 64. | The value must be equal to or less than 256. |
 
 ## `ipc_endpoint` section