Benchmark: Don't manually build Lotus

[LesnyRumcajs]

Issue summary

There is no need to manually build Lotus from the source unless we change some flags or node settings. We can use docker images that the Lotus team already regularly publishes.

The stable releases are available here.

For mainnet we should use stable tag, for calibnet it will be stable-calibnet. We should not use master like in the current scripts.

This should cut benchmark times, and reduce the dependencies hell (right now, if new Lotus would introduce a new dependency, the benchmarks would fail).

Acceptance criteria

• Benchmark doesn't build Lotus, it is using Lotus docker images,
• Lotus versions are stable, not development,
• Dependencies are cleaned
• Documentation (if any) is updated

Other information and links