fix(tracing): clean up processors after tracing is disabled

[sylvesterkaczmarek]

Summary

• always run DefaultTraceProvider.shutdown() for registered processors
• keep shutdown independent from whether new tracing is currently disabled
• ensure processors created before or while tracing is disabled can release worker and client resources

DefaultTraceProvider.shutdown() previously returned early when the provider's disabled flag was set. Disabling tracing controls creation of new trace data, but registered processors can already own resources and still require shutdown. This is especially relevant for the default batch processor, which owns a worker thread and exporter client.

Test plan

• added focused regression coverage in tests/test_disabled_trace_provider_shutdown.py
• GitHub Actions

Issue number

N/A