[1.2 regression] Math on multiple selectors returns distinct points

[kostasb]

v1.2.1 behavior:

> insert metrics value=1
> insert metrics value=2
> SELECT min(value),max(value) from metrics
name: metrics
time min max
---- --- ---
0    1   2

> SELECT min(value)*1,max(value)*1 from metrics
name: metrics
time                min max
----                --- ---
1490013334977108401 1   
1490013337400734540     2

1.1.4 behavior:

> insert metrics value=1
> insert metrics value=2
> SELECT min(value),max(value) from metrics
name: metrics
time	min	max
----	---	---
0	1	2

> SELECT min(value)*1,max(value)*1 from metrics
name: metrics
time	min	max
----	---	---
0	1	2

May be related to #8072

@jsternberg

[DAfanasyev]

Looks like it is third issue about the same regression:
#8164

[jsternberg]

Bisect says it was introduced by subqueries. I'll be looking into this. I'll close the other issues to consolidate around this one.

Bisect says it's this commit, but it's a large one: d7c8c7c.

[jsternberg]

If we don't backport a fix for this, you can use subqueries as a workaround.

> select max*1, min*1 from (select max(value), min(value) from metrics)
name: metrics
time max min
---- --- ---
0    2   1

I seem to have isolated the issue. @jwilder should we backport this fix?

[timhallinflux]

@rkuchan -- please add this to the release notes for 1.2.2 enterprise as a known issue -- along with the proposed workaround.