Add Sentry Integration for Error Monitoring in akavesdk-py

[gitsofaryan]

Summary

Introduce optional Sentry-based error monitoring and observability support in akavesdk-py to improve production debugging, failure visibility, and operational insight without introducing a hard dependency.

---

Problem

Currently, akavesdk-py does not provide centralized error monitoring or structured runtime observability.

When SDK users encounter issues such as:

• Network or transport failures
• O3 streaming errors
• CID resolution failures
• Authentication / UCAN validation errors
• Retry exhaustion from tenacity

There is no built-in mechanism to:

• Capture stack traces centrally
• Monitor error frequency
• Attach contextual metadata to failures
• Trace performance bottlenecks
• Debug real-world production issues

This limits visibility for both SDK maintainers and integrators.

---

Proposed Solution

Integrate Sentry (Python SDK) as an optional observability layer that:

• Captures unhandled exceptions
• Captures structured SDK-level errors
• Tracks retry exhaustion events
• Attaches contextual metadata (dataset_id, CID, node endpoint, operation name)
• Supports async workflows
• Optionally enables performance tracing

The integration must:

• Be optional (no forced dependency)
• Be configurable via environment variables
• Avoid logging sensitive information (private keys, tokens, auth headers)
• Introduce minimal performance overhead
• Not create breaking changes

---

High-Level Implementation Plan

1. Add Sentry as an optional dependency (akavesdk[sentry])
2. Create a centralized observability module inside the SDK
3. Provide an initialization function (e.g., init_sentry)
4. Instrument core SDK error boundaries
5. Capture retry exhaustion events from tenacity
6. Optionally add tracing spans around streaming/upload operations
7. Add documentation explaining setup and configuration

---

Security Considerations

• Ensure private keys and UCAN tokens are never logged
• Mask authentication headers
• Avoid sending large payload bodies
• Provide a sanitization hook if needed

---

Expected Impact

• Production-grade observability
• Faster debugging for SDK consumers
• Better insight into decentralized network reliability
• Improved SDK trust and enterprise readiness
• Clearer operational visibility during distributed ML workloads

---

Acceptance Criteria

• Optional dependency added
• Observability module implemented
• Core error boundaries instrumented
• Retry exhaustion captured
• Documentation updated
• No breaking changes introduced