Stored timestamps in `_value` interpreted as strings in Flux query

Nobody has claimed this yet.

When applying a conversion from time to int or uint in e2e tests we get the following error:

```
--- FAIL: TestFluxEndToEnd/to_int.flux (0.23s)
end_to_end_test.go:170: failed to read meta data: failed to evaluate map function: strconv.ParseInt: parsing "1970-01-01T00:00:00.000000000Z": invalid syntax

--- FAIL: TestFluxEndToEnd/to_uint.flux (0.22s)
end_to_end_test.go:162: failed to evaluate map function: strconv.ParseUint: parsing "1970-01-01T00:00:00.000000000Z": invalid syntax
```

The table that causes the error is this one:

```
#datatype,string,long,dateTime:RFC3339,dateTime:RFC3339,string,string
#group,false,false,false,false,true,true
#default,_result,,,,,
,result,table,_time,_value,_field,_measurement
,,4,2018-05-22T19:53:40Z,1970-01-01T00:00:00Z,k4,m
,,4,2018-05-22T19:53:41Z,2018-05-22T19:53:26.033Z,k4,m
,,4,2018-05-22T19:53:42Z,2018-05-22T19:53:26.033066Z,k4,m
,,4,2018-05-22T19:53:43Z,2018-05-22T19:00:00+01:00,k4,m
```

The error happens here https://github.com/influxdata/flux/blob/52bf9ec1b0e1b0be208f111b10d2d4638ea8d23c/stdlib/universe/typeconv.go#L183 (for ints).

It seems that the value gets erroneously interpreted as a string instead of a timestamp.

The problem doesn't happen in Flux, so I suspect that this has something to do with the fact that, in InfluxDB, we store data using a `to` transformation before querying. Maybe `to` has some problems in storing `_value` as a `dateTime:RFC3339`.

Start with stdlib/universe/typeconv.go at the referenced conversion and trace how stored _value timestamps reach the Flux type conversion. Reproduce with end_to_end_test.go using to_int.flux and to_uint.flux; done means the timestamp-valued table no longer produces string-parsing errors during those conversions.

Written by the indexing model from the issue text.