Rename max_write_buffer_size in _ccr/stats as it's misleading

[dliappis]

The max_write_buffer_size parameter name is misleading: it gives the impression that you are controlling directly the size of the allocated write buffer per shard, however it is a threshold controlling the amount of subsequent reads.

For example with the following index settings:

{
  "max_concurrent_read_batches": 1,
  "max_concurrent_write_batches": 1,
  "max_batch_operation_count": 32768,
  "max_batch_size_in_bytes": 4718592,
  "max_write_buffer_size": 33000
}

it's possible that one read brings max_batch_operation_count 32768 # of docs, the subsequent read sees that max_write_buffer_size is still higher (33000) and retrieves another batch of 32768 documents. In this case and instance, the _ccr/stats field number_of_queued_writes correctly reports 65535.

One suggestion would be to rename:
max_write_buffer_size -> max_subsequent_reads_into_write_buffer or replace subsequent with next to make it shorter. Any other suggestion is welcome of course.

[elasticmachine]

Pinging @elastic/es-distributed

[dliappis]

cc @jasontedor for awareness.

[bleskes]

+1 to rename. I would propose max_queued_writes

[dliappis]

@martijnvg As #34797 already renames max_write_buffer_size to max_write_buffer_count, I guess we can close #33906?

[martijnvg]

@dliappis Yes, I think we can close this issue.

[dliappis]

Closed as per #33906 (comment)