[YSQL] YSQL reads all the columns from DocDB when subset is enough

[d-uspenskiy]

For processing some of the SQL queries which uses subset of columns in their result YSQL reads all the columns from DocDB. Reading only necessary columns will reduce network traffic and improve the performance.

Example:

CREATE TABLE t(k INT PRIMARY KEY, short_data INT, long_data TEXT);
INSERT INTO t values(1, 1, 'long text');
SELECT * FROM t WHERE k = 1;
SELECT k FROM t WHERE k = 1;
SELECT short_data FROM t WHERE k = 1;
SELECT long_data FROM t WHERE k = 1;

All 4 selects sends same requests to tserver (with reading all 3 columns + ybctid)

stmt_id: 94082278811152
schema_version: 0
partition_column_values {
  value {
    int32_value: 1
  }
}
targets {
  column_id: 10
}
targets {
  column_id: 11
}
targets {
  column_id: 12
}
targets {
  column_id: -8
}
column_refs {
  ids: 10
  ids: 11
  ids: 12
}
is_forward_scan: true
is_aggregate: false
limit: 1024
return_paging_state: true
ysql_catalog_version: 2
table_id: "000030af00003000800000000000400a"

Note: queries without WHERE clause sends read request with only necessary columns.