feat: memory profiler

[alv-around]

Hi,

Here is the PR for #1029 which implements metrics recorder for Prometheus. Previous discussion can be found in #1201.

The PR implements a recoder that:

1. Ttarts a thread which collects process-metrics for a given Duration. It is important that the process metrics (memory/cpu usage.. etc) in it's own separate thread spawned is, otherwise, in an async task will be block anytime the VM performs any of it's cpu-intesive computations.
2. [metrics_context_tracing](https://docs.rs/metrics-tracing-context/latest/metrics_tracing_context/) enabled( as stated in the docs). so any metric recorded within a span, the span is automatically added as a metric label. However, this does not apply to the process metrics metrics regularly collected in the previous point, because they run in a separate thread.

Usage: In this repo, I made two examples how the memory profiler could work with both (pull/push) prometheus exporters.

[jonathanpwang]

Sorry for delay, I'll try to get to this soon.

[alv-around]

Sorry for delay, I'll try to get to this soon.

Sure no problem, I rebased the branched to latest main.

EDIT:
Nevermind about the rebase. That totally went the wrong way. Looking forward for your feedback.