fix bug in lake scan when merging by non-ts pool key

[mccanne]

closes #2749

[nwt]

We're eventually going to have to address the fact that zbuf.MergeByTs doesn't do the right thing for records with a non-time ts field.

[mccanne]

I think this can easily be part of the JIT-CG. The scan scheduler can look at the metadata across a partition to see inspect the type signature(s) for the pool key.

[philrz]

FWIW, I just tested out the branch using the original repro steps and it looks good to me too.

$ zed -version
Version: v0.29.0-350-gce2003e6

$ rm -rf $ZED_LAKE_ROOT

$ zed lake init
using environment variable ZED_LAKE_ROOT
lake created: file:///Users/phil/logs

$ zed lake create -S 5MB -p logs -orderby id.resp_h:asc
pool created: logs

$ zed lake load -p logs ~/work/zed-sample-data/zeek-default/conn.log.gz
1sozz07rJ7a6VZ3SIRwdLALOJGj committed 17 segments

$ zed lake query -z 'from logs | cut id.resp_h'
{id:{resp_h:2.22.230.64}}
{id:{resp_h:2.22.230.64}}
{id:{resp_h:2.22.230.64}}
{id:{resp_h:4.53.160.75}}
{id:{resp_h:4.53.160.75}}
{id:{resp_h:4.53.160.75}}
{id:{resp_h:4.53.160.75}}
{id:{resp_h:5.9.78.71}}
{id:{resp_h:5.9.102.20}}
...

$ zed lake query -z 'from logs | cut id.resp_h' > /tmp/ip-query

$ zed lake query -z 'from logs | cut id.resp_h | sort' > /tmp/ip-query-sorted

$ diff /tmp/ip-query /tmp/ip-query-sorted
$ echo $?
0