flux: Extremely inefficient when querying the first/last entry

Nobody has claimed this yet.

__Steps to reproduce:__
List the minimal actions needed to reproduce the behavior.

1. create a measurement with a lot of entries

- in total 10000000 entries over 3 year
- 5000000 in this time period are matching the a="b" where clause

2. query last or first entry for measurement - system has extreme high CPU load

__Expected behavior:__

Works as fast as InfluxQL:

# first entry
select x,y,z from foobar where a='b' order by time asc limit 1;

# last entry
select x,y,z from foobar where a='b' order by time desc limit 1;

__Actual behavior:__

A query for the last/first entry require a start range of 0 because it is not previously know. So the whole system spikes in CPU and memory usage and at the end blocks everything.

# first entry
from(bucket:"mybucket")
|> range(start:0)
|> filter(fn: (r) =>
r._measurement == "foobar" and
(r._field == "x" or
r._field == "y" or
r._field == "z") and
r.a == "b")
|> limit(n:1)
|> pivot(rowKey:["_time"], columnKey: ["_field"], valueColumn: "_value")
|> group()
|> sort(columns:["_time"], desc:false)
|> limit(n:1)

# last entry
from(bucket:"mybucket")
|> range(start:0)
|> filter(fn: (r) =>
r._measurement == "foobar" and
(r._field == "x" or
r._field == "y" or
r._field == "z") and
r.a == "b")
|> tail(n:1)
|> pivot(rowKey:["_time"], columnKey: ["_field"], valueColumn: "_value")
|> group()
|> sort(columns:["_time"], desc:false)
|> tail(n:1)

__Environment info:__

* System info: Linux 4.19.0-13-amd64 x86_64
* InfluxDB version: InfluxDB v1.8.3 (git: 1.8 563e6c3d1a7a2790763c6289501095dbec19244e)

__Config:__

[http]
flux-enabled = true

Start by reproducing the Flux queries against a measurement with the described large dataset and compare first/last behavior with the InfluxQL examples. Trace the query path for range(start:0), limit, and tail; done means first/last queries complete without the reported extreme CPU and memory usage.

Written by the indexing model from the issue text.