fix(opentelemetry): recreate tracer object after plugin metadata changed

[shreemaan-abhishek]

Description

The opentelemetry plugin caches the tracer object per route via core.lrucache.plugin_ctx, but passed nil as the cache version key in _M.rewrite. The tracer bakes in everything sourced from plugin_metadata (collector address, request_timeout/request_headers, trace_id_source, sampler, and resource attributes). With a nil version key the cache entry is never invalidated, so after an operator updates the opentelemetry plugin_metadata at runtime (e.g. points to a new OTLP collector or flips trace_id_source), the worker keeps serving the stale cached tracer. Spans keep going to the old endpoint until the per-conf cache entry is evicted or the worker recycles.

This passes metadata.modifiedIndex as the lrucache version key so the tracer is rebuilt whenever the plugin_metadata changes. api_ctx.conf_version (the existing version) only tracks the route/service, not plugin_metadata, so a metadata-only change does not invalidate the entry on its own.

Fixes the stale-tracer behavior on plugin_metadata change.

Checklist

• I have explained the need for this PR and the problem it solves
• I have explained the changes or the new features added to this PR
• I have added tests corresponding to this change
• I have updated the documentation to reflect this change (N/A, behavior-internal)
• I have verified that this change is backward compatible (If not, please discuss on the APISIX mailing list first)