Add supplementary guidelines for SDK self-observability

[cijothomas]

Follow-up to #5087, which introduced specification/self-observability.md.

The current document is intentionally minimal. Based on feedback from the spec community meeting (2026-06-02), implementations would benefit from non-normative supplementary guidelines covering:

• Lifecycle and ordering — how to handle telemetry emitted before the recording provider is initialized (buffering, throwaway/bootstrap loggers) and the symmetric shutdown case.
• Avoiding telemetry-induced-telemetry loops — recommended patterns to prevent the SDK from recursively instrumenting its own export pipeline.
• Startup configuration event — emitting a single verbose event/log on SDK init with the resolved effective configuration (env vars, exporters, processors, detected resource) to aid debugging.
• Scope of signals — clarifying that self-observability is metrics-focused today but may extend to logs/traces; designing for that.
• Telemetry versioning — guidance on schema URLs and handling stability transitions for self-observability telemetry.
• Opt-in / performance — patterns used by existing SDKs (e.g., Go's OTEL_GO_X_OBSERVABILITY, Java's configuration toggle).

Proposed location: specification/self-observability-supplementary-guidelines.md, following the pattern of specification/{logs,metrics,configuration}/supplementary-guidelines.md.