[Disk Manager] Tracing with Open Telemetry

[gy2411]

Traced should be collected for all requests to Disk Manager should be traced. Given request_id, we need opportunity to see what was this request doing (which tasks scheduled, which queries to nbs/nfs/ydb/s3 made, when exactly it happened and how much time did different steps take).

We are going to use Open Telemetry framework https://opentelemetry.io/. Namely, we should add instrumentation to Disk Manager code for creating spans and events. Also Disk Manager must be able to use trace context passed by caller (like trace id and parent span id) if it is present. Also Disk Manager must be able to export traces to a backend (in order to store and visualize traces).

[gy2411]

Further improvements of tracing in Disk Manager:

• Do not sample spans for large task generation
• Handle retriable errors in spans properly (decide should we set Erorr status for such spans or not).
• Shorten error message of non-retryable errors in spans (do not write stacktrace, do not write retriable errors from errors chain).
• Add request id attribute to spans of tasks. Also think should we add any other attributes, events for error handling to spans of tasks.
• Add events for retriable errors in blockstore go SDK.