100% cpu consummatiun from 2.07 to 2.0.8

Hi, this bug is non blocking (for me) because I work around adding CPU.

__Steps to reproduce:__
1. install database influx 2.0.7, put about 70Go of time series data on a virtual machine with only ONE processor
2. inject about 1Mo of data (1Mo is the size of the text message) each 10s
3. the cpu is about 25%
4. upgrade to 2.0.8 and restart
5. the CPU is now about 100%, request are very slow
6. change the virtual machine setting from 1 processor to 4 processor, and ram from 4GB to 8GB
7. the service return to initial CPU consummation (divide by 4 because 4 process)

__Expected behavior:__
The CPU consummation doesn't change after upgrade.
Note the ram doesn't

__Actual behavior:__
the CPU is now about 100%, request are very slow.

__Environment info:__

* System info: Linux 4.18.0-305.12.1.el8_4.x86_64 x86_64
* InfluxDB version:
* 2.0.7
* InfluxDB 2.0.8 (git: e91d41810f) build_date: 2021-08-13T18:22:30Z
* Other relevant environment details:
* 1CPU and them 4CPU.
* note influxdb2 eat 79 330 176 kB (77GB) of VSZ (!)

__Performance:__
I can't downgrade the production to run performance measurement.
I attach a commented measurement of the computer metric (provided by the influxdb2 database itself)

Start by comparing InfluxDB 2.0.7 and 2.0.8 under the reported one-CPU Linux environment, using the attached computer-metric measurement as the baseline. Reproduce the 100% CPU and slow requests with the stated data volume, then identify the regression and verify that CPU use and request performance return to the 2.0.7 behavior.

Written by the indexing model from the issue text.