Currently there are two forms for running a retryable bulk write. In the first, the command specifies no statement ids at all, and the command execution assigns statement IDs implicitly to each entry in the bulk write. In the second, the command specifies the stmtId for each write in the bulk explicitly, in a second 'stmtIds' array that maps 1-1 to the array of write operations in the batch. This ticket is for adding a third form, where there is a single top-level stmtId specified to the write, and the server assigns that stmtId to the first write in the batch, then assigns new stmtIds to each additional write in the batch implicitly.