Unify http_listener error response with influxdb

[glinton]

Resolves #4742

[danielnelson]

The error strings need to match up with what the influxdb output expects to trigger the cases that will cause an upstream Telegraf to retry.

telegraf/plugins/outputs/influxdb/http.go

Lines 320 to 335 in af0ef55

	// Other partial write errors, such as "field type conflict", are not
	// correctable at this point and so the point is dropped instead of
	// retrying.
	if strings.Contains(desc, errStringPartialWrite) {
	log.Printf("E! [outputs.influxdb]: when writing to [%s]: received error %v; discarding points",
	c.URL(), desc)
	return nil
	}
	// This error indicates a bug in either Telegraf line protocol
	// serialization, retries would not be successful.
	if strings.Contains(desc, errStringUnableToParse) {
	log.Printf("E! [outputs.influxdb]: when writing to [%s]: received error %v; discarding points",
	c.URL(), desc)
	return nil
	}

[glinton]

That seems quite brittle, but good point. This depends on #4743 as the errors returned from the parser doesn't return the proper errors:

# telegraf
$ curl -XPOST 'http://telegraf:8086/write' --data-binary 'cpu value='
{"error":"metric parse error: expected field at offset 10: \"cpu value=\""}

$ echo -e "cpu value=84\ncpu value=" | curl 'http://telegraf:8086/write?db=telegraf' --data-binary @-
{"error":"metric parse error: expected field at offset 23: \"cpu value=84\\ncpu value=\\n\""}


# influx
$ curl -i -XPOST 'http://influx:8086/write?db=telegraf' --data-binary 'cpu value='
{"error":"unable to parse 'cpu value=': missing field value"}

$ echo -e "cpu value=84\ncpu value=" | curl -i 'http://influx:8086/write?db=telegraf' --data-binary @-
{"error":"partial write: unable to parse 'cpu value=': missing field value dropped=0"}

Should we also require a db to be set? Writing to influx at /write returns {"error":"database is required"}.