Fix leaks in 'trace' element

[denis-tingaikin]

https://github.com/networkservicemesh/sdk/commit/15fe129fec59610b4da3b4887f2d7b29d4705d9f

Steps to reproduce

It's actual for metrics and for logger spans (when otel is disabled)

1. Run https://github.com/networkservicemesh/deployments-k8s/tree/main/examples/observability/jaeger_and_prometheus
2. Continus scale client/endpoint
3. Check mem profiles

Actual: Memory leaks. Expected: No memory leaks.

Action plan: (provided by @VitalyGushin )

• [x] Analysis of existing profiles from the community
• [x] Scaling test rc1
• [x] Scaling test rc1 + tracing enabled, compare results
• [x] Code research (TELEMETRY=false scenario) 7h
• [x] Prepare fixes and execute scaling test for logrus/opa/recvfd components 14h
• [x] Prepare builds for scaling test rc3 + tracing enabled, test/result analysis 3h
• [x] Identifying memory leaking methods in profiles 3h
• [x] Provide testing/analytics for the TELEMETRY=true scenario 4h
• [ ] Prepare fixes and check them with benchmark tests 14h
• [ ] Prepare images, perform the scaling test with profiling, results analysis 4h
• [ ] Prepare PR (in case of a successful result in the profile), review fixes 3h

[denis-tingaikin]

@VitalyGushin profiles:

forwarder-vpp_memprofiles_20240519135433.tar (1).gz nsmgr_memprofiles_20240519135233.tar (1).gz

[denis-tingaikin]

Finally, we've found a root cause with the trace element, and it will be considered in the separate issue https://github.com/networkservicemesh/sdk/issues/1667