refactor(sdk): change SpanProcessor::on_end to take &mut FinishedSpan

[bryantbiggs]

Summary

Supersedes #2962 (original work by @paullegranddc). Refactors SpanProcessor::on_end from fn on_end(&self, span: SpanData) to fn on_end(&self, span: &mut FinishedSpan), enabling zero-copy span inspection and ownership semantics for multi-processor chains.

Breaking Change

SpanProcessor::on_end now takes &mut FinishedSpan instead of SpanData.

Key Changes

• ReadableSpan trait: New trait implemented by both Span (live, for on_start) and FinishedSpan (ended, for on_end). Allows processors to inspect span fields (name(), attributes(), span_kind(), instrumentation_scope(), etc.) without cloning.
• FinishedSpan type: Wraps finished span data with consume semantics. consume() gives the last processor zero-copy ownership via take(); earlier processors receive a clone(). is_consumed() lets processors check post-consume state.
• end_time_set flag: Replaces the fragile start_time == end_time sentinel comparison. end_with_timestamp() now correctly preserves user-provided timestamps even when equal to start time.
• Zero-copy span_context: Uses mem::replace instead of clone() in end_and_export to avoid heap allocation for TraceState.
• end_and_export rewrite: Takes span data before shutdown check (fixes potential data leak on re-entrant Drop), sets end time via explicit flag.

Migration Guide

// Before
impl SpanProcessor for MyProcessor {
    fn on_end(&self, span: SpanData) {
        export(span);
    }
}

// After — read without cloning
impl SpanProcessor for MyProcessor {
    fn on_end(&self, span: &mut FinishedSpan) {
        if span.name() == Some("skip_me") { return; }
        if let Some(data) = span.consume() {
            export(data);
        }
    }
}

Files Changed

• opentelemetry-sdk/src/trace/span.rs — FinishedSpan, ReadableSpan trait, end_and_export rewrite, end_time_set flag
• opentelemetry-sdk/src/trace/span_processor.rs — Updated SpanProcessor trait + built-in impls
• opentelemetry-sdk/src/trace/span_processor_with_async_runtime.rs — Async batch processor updated
• opentelemetry-sdk/src/trace/tracer.rs — end_time_set: false in span construction
• examples/tracing-http-propagator/src/server.rs — New RouteConcurrencyCounterSpanProcessor example demonstrating ReadableSpan in both on_start and on_end
• opentelemetry-sdk/benches/span_processor_api.rs — New benchmark (0/1/2/4 processors, ~191 ns uniform on M4 Max)
• stress/src/traces.rs — Updated NoOpSpanProcessor

Benchmark Results (Apple M4 Max)

Processors	Time
0	190.82 ns
1	192.53 ns
2	191.67 ns
4	193.48 ns

Uniform across processor counts — the new API adds negligible overhead.

Test Plan

• All existing span/span_processor tests pass (47 lib tests + 38 doc tests)
• New tests: test_readable_span_after_consume_returns_defaults, test_end_with_timestamp_equal_to_start_time
• clippy clean, fmt clean
• Example (tracing-http-propagator) and stress test compile
• Async runtime (rt-tokio) compiles
• Benchmarks compile and run

Closes #2726, #2939. Relates to #2940.

[codecov]

Codecov Report

❌ Patch coverage is 83.14917% with 61 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.8%. Comparing base (ec289cb) to head (f2c4e7b).

Files with missing lines	Patch %	Lines
opentelemetry-sdk/src/trace/span.rs	81.4%	60 Missing ⚠️
opentelemetry-sdk/src/trace/span_processor.rs	95.8%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #3410     +/-   ##
=======================================
- Coverage   82.8%   82.8%   -0.1%     
=======================================
  Files        130     130             
  Lines      27289   27595    +306     
=======================================
+ Hits       22622   22865    +243     
- Misses      4667    4730     +63     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[bantonsson]

Thanks for moving this forward @bryantbiggs. Only some minor things.

[bantonsson]

Can we get some maintainer 👀 on this?
/cc @cijothomas @hdost @lalitb @utpilla @TommyCpp

[bantonsson]

Any of the maintainers have some time for this?
/cc @cijothomas @hdost @lalitb @utpilla @TommyCpp

[cijothomas]

Thanks for the PR! Let's hold off on this one until after the current release ships — that way we can look at it together with the other pending tracing changes and batch all the breaking changes into the next release. Will pick it up right after — appreciate the patience!