refine IndexLookUp executor runtime stats

[crazycs520]

Is your feature request related to a problem? Please describe:

Currently, the runtime stats of IndexLookUp is hard to understand, here is an example:

test> explain analyze select a+b from t1 use index (a) where a<100000 order by b limit 500;
+--------------------------------+----------+---------+-----------+----------------------+-------------------------------------------------------------------------------------------------------------+---------------------------------------+-------------------+------+
| id                             | estRows  | actRows | task      | access object        | execution info                                                                                              | operator info                         | memory            | disk |
+--------------------------------+----------+---------+-----------+----------------------+-------------------------------------------------------------------------------------------------------------+---------------------------------------+-------------------+------+
| Projection_9                   | 500.00   | 500     | root      |                      | time:145.126133ms, loops:2, Concurrency:OFF                                                                 | plus(test.t1.a, test.t1.b)->Column#4  | 8.34375 KB        | N/A  |
| └─TopN_11                      | 500.00   | 500     | root      |                      | time:145.112393ms, loops:2                                                                                  | test.t1.b, offset:0, count:500        | 12.25 KB          | N/A  |
|   └─IndexLookUp_19             | 500.00   | 50007   | root      |                      | time:144.250313ms, loops:51, cop_task: {num: 1, max:371.69µs, proc_keys: 0, rpc_num: 1, rpc_time: 363.18µs} |                                       | 1.529541015625 MB | N/A  |
|     ├─IndexRangeScan_16(Build) | 50021.28 | 50007   | cop[tikv] | table:t1, index:a(a) | time:21ms, loops:53                                                                                         | range:[-inf,100000), keep order:false | N/A               | N/A  |
|     └─TopN_18(Probe)           | 500.00   | 0       | cop[tikv] |                      | time:0ns, loops:0                                                                                           | test.t1.b, offset:0, count:500        | N/A               | N/A  |
|       └─TableRowIDScan_17      | 50021.28 | 0       | cop[tikv] | table:t1             | time:0ns, loops:0                                                                                           | keep order:false                      | N/A               | N/A  |
+--------------------------------+----------+---------+-----------+----------------------+-------------------------------------------------------------------------------------------------------------+---------------------------------------+-------------------+------+

there is some problem:

1.IndexLookUp reader uses IndexScan to read index first, and got the handle value from index value, then use TableRowIDScan to read data. So, we need to know the information such as below:

• RPC runtime stats of IndexScan
• RPC runtime stats of TableRowIDScan

Describe the feature you'd like:

Here is an example of IndexLookUp runtime stats:

time:300ms, loops:501, index_task: {num: 1, time: 10ms, rpc_num: 2, rpc_time: 8ms}, table_task: {num: 503, time: 290ms, rpc_num: 509, rpc_time: 8ms}, backoff: {tikvRPC: 10ms, regionMiss: 30ms}

[crazycs520]

@qw4990 @zz-jason PTAL

[jyz0309]

Alkaid will get it done:>