fatal error: concurrent map read and map write with opentsdb?

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

1. I don't know how to reproduce the issue consistently. It appears to be random.

__Expected behavior:__
No crash

__Actual behavior:__
InfluxDB crashes about once a week (but recovers automatically).

__Environment info:__
There are two influxdb instances running on two identical ppc64le nodes.
Collectd sends data from multiple hosts via the `write_opentsdb` plugin to both instances.
* System info: `Linux 4.14.0-115.14.1.el7a.ppc64le ppc64le`, `RHEL 7.6 for ppc64le`
* InfluxDB version: `InfluxDB v1.7.10 (git: heads/v1.7.10 f46f63d4e2d9684a2dd716594ab609ccd32f0a5b)`
* Other relevant environment details: Bare metal, built with Go `1.12.17` and:
```
go get github.com/influxdata/influxdb
cd $GOPATH/src/github.com/influxdata/influxdb
git checkout tags/v1.7.10 -b v1.7.10

python build.py --package --release --clean --update
```
__Config:__
Non default config entries:
```
reporting-disabled = true

[data]
index-version = "tsi1"
query-log-enabled = false
series-id-set-cache-size = 100

[http]
auth-enabled = true
log-enabled = false

[[opentsdb]]
enabled = true
bind-address = ":4242"
database = "esmon_database"

[continuous_queries]
log-enabled = false
```

__Logs:__
[influxdb_crash_node1.log](https://github.com/influxdata/influxdb/files/4343088/influxdb_crash_node1.log)
[influxdb_crash_node2.log](https://github.com/influxdata/influxdb/files/4343089/influxdb_crash_node2.log)

Start by reviewing influxdb_crash_node1.log and influxdb_crash_node2.log, then trace the enabled [[opentsdb]] input and its concurrent write path. Done means identifying the race and confirming that concurrent OpenTSDB ingestion no longer crashes InfluxDB.

Written by the indexing model from the issue text.