[WIP] Fix avg_bytes_per_rows when node 0 does not have data

[Christian8491]

No description provided.

[wmalpica]

I think we should always use that info, not only if the master node does not have it.
Second, i think you should weight the avg_bytes_per_row info by the amount of data from that node, such that if one node has little data but has a very high or low avg_bytes_per_row, it does not skew the results. So I would recommend doing something like:

size_t total_bytes = 0; 
size_t total_rows = 0;
for(std::size_t i = 0; i < nodes.size(); ++i) {
total_bytes += total_avg_bytes_per_row[i]*total_table_rows[i];
total_rows += total_table_rows[i];
}
size_t final_avg_bytes_per_row = total_bytes/total_rows;

[wmalpica]

and use final_avg_bytes_per_row for generate_partition_plan

[Christian8491]

updated as requested.