Flux doesn't respect coordinator limits

Flux doesn't appear to be respecting configured [coordinator] limits.

Steps to reproduce: Create Grafana data-source on some real database and click test.
It sends query:
from(bucket:"db/rp")
|> range(start:-10y)
|> last()
In a couple of minutes InfluxDB is Out Of Memory and crashed.

InfluxDB 1.8

[coordinator]
query-timeout = "30s"
log-queries-after = "10s"
max-select-point = 4194304
max-select-series = 65536
max-select-buckets = 16384

Start by reproducing the Grafana data-source test against InfluxDB 1.8 with the shown coordinator settings and query. Trace how Flux handles coordinator limits for the long-range last() query; done means the configured limits are respected and the query no longer causes InfluxDB to run out of memory and crash.

Written by the indexing model from the issue text.