Typesense uses host machine memory stats when running in Docker

[b0g3r]

Description

Typesense uses available memory stats to report it on typesense_memory_* metrics and for throttling writes with memory-used-max-percentage settings. But in docker container, where resource limits are set via cgroups, typesense reads the available memory stats from the host machine. It leads to OOM kills even with set up memory-used-max-percentage and incorrect metrics reporting.

The same issue is for Kubernetes environment with resources:limits: where nodes are usually huge servers.

Side note: it is also true for CPU count which is used to calculate default values for some settings, but IMO it is not so important.

Steps to reproduce

https://gist.github.com/b0g3r/2fcf88772cddf4bd4630a48bc1d63073