perf: Stop holding the global mutex during scope access

[jpnurmi]

Make sentry-native multi-threading friendly by allowing concurrent scope data readers.

The global scope data mutex became a serious bottleneck for multi-threaded logs and metrics. Most notably, when wired up with UE_LOG that may capture Unreal Engine logs from 100+ threads.

The previous step (#2042) enabled RW locks on the scope data. This final step removes the remaining, now superfluous global mutex around the data. With this, multiple logger threads can enrich incoming log items with scope data/attributes in parallel.

In theory, other concurrent scope operations may benefit too, but logs and metrics are the most important high-volume scenario.

Before

tests/benchmark.py::test_benchmark_logs[1] PASSED
Min 0.007ms, Max 0.007ms, Mean 0.007ms, Median 0.007ms, CPU 0.007ms
tests/benchmark.py::test_benchmark_logs[8] PASSED
Min 0.037ms, Max 0.037ms, Mean 0.037ms, Median 0.037ms, CPU 0.012ms
tests/benchmark.py::test_benchmark_logs[16] PASSED
Min 0.067ms, Max 0.067ms, Mean 0.067ms, Median 0.067ms, CPU 0.008ms
tests/benchmark.py::test_benchmark_logs[32] PASSED
Min 0.137ms, Max 0.137ms, Mean 0.137ms, Median 0.137ms, CPU 0.008ms
tests/benchmark.py::test_benchmark_metrics[1] PASSED
Min 0.004ms, Max 0.004ms, Mean 0.004ms, Median 0.004ms, CPU 0.003ms
tests/benchmark.py::test_benchmark_metrics[8] PASSED
Min 0.033ms, Max 0.033ms, Mean 0.033ms, Median 0.033ms, CPU 0.007ms
tests/benchmark.py::test_benchmark_metrics[16] PASSED
Min 0.059ms, Max 0.059ms, Mean 0.059ms, Median 0.059ms, CPU 0.007ms
tests/benchmark.py::test_benchmark_metrics[32] PASSED
Min 0.126ms, Max 0.126ms, Mean 0.126ms, Median 0.126ms, CPU 0.007ms

After

tests/benchmark.py::test_benchmark_logs[1] PASSED
Min 0.005ms, Max 0.005ms, Mean 0.005ms, Median 0.005ms, CPU 0.004ms
tests/benchmark.py::test_benchmark_logs[8] PASSED
Min 0.013ms, Max 0.013ms, Mean 0.013ms, Median 0.013ms, CPU 0.012ms
tests/benchmark.py::test_benchmark_logs[16] PASSED
Min 0.014ms, Max 0.014ms, Mean 0.014ms, Median 0.014ms, CPU 0.012ms
tests/benchmark.py::test_benchmark_logs[32] PASSED
Min 0.016ms, Max 0.016ms, Mean 0.016ms, Median 0.016ms, CPU 0.008ms
tests/benchmark.py::test_benchmark_metrics[1] PASSED
Min 0.005ms, Max 0.005ms, Mean 0.005ms, Median 0.005ms, CPU 0.005ms
tests/benchmark.py::test_benchmark_metrics[8] PASSED
Min 0.011ms, Max 0.011ms, Mean 0.011ms, Median 0.011ms, CPU 0.010ms
tests/benchmark.py::test_benchmark_metrics[16] PASSED
Min 0.017ms, Max 0.017ms, Mean 0.017ms, Median 0.017ms, CPU 0.015ms
tests/benchmark.py::test_benchmark_metrics[32] PASSED
Min 0.034ms, Max 0.034ms, Mean 0.034ms, Median 0.034ms, CPU 0.020ms

Close: #1862

[cursor]

Stale Bugbot comment from a previous run.

[codecov]

Codecov Report

❌ Patch coverage is 94.44444% with 11 lines in your changes missing coverage. Please review.
✅ Project coverage is 75.21%. Comparing base (233ea8b) to head (34af03d).

Additional details and impacted files

@@                   Coverage Diff                    @@
##           jpnurmi/perf/rwscope    #2091      +/-   ##
========================================================
+ Coverage                 74.47%   75.21%   +0.73%     
========================================================
  Files                       100      103       +3     
  Lines                     27403    27930     +527     
  Branches                   4968     5094     +126     
========================================================
+ Hits                      20408    21007     +599     
+ Misses                     5724     5574     -150     
- Partials                   1271     1349      +78     

🚀 New features to boost your workflow:

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

[cursor]

Stale Bugbot comment from a previous run.

[limbonaut]

Caught a race: two threads can update the same tag, for example, but their notifications can arrive in the opposite order. An observer can therefore end up remembering an older value instead of the value actually stored in the scope. See comments below for more details.

[limbonaut]

Sweet! Benchmarks with the recent changes:

Median benchmark-reported real time, in milliseconds.

Benchmark	Threads	Master	PR	Speedup
Logs	1	0.00332	0.00348	0.96×
Logs	8	0.01571	0.00619	2.54×
Logs	16	0.04656	0.00917	5.08×
Logs	32	0.14999	0.02016	7.44×
Metrics	1	0.00266	0.00279	0.95×
Metrics	8	0.02256	0.00552	4.09×
Metrics	16	0.05678	0.00922	6.16×
Metrics	32	0.13757	0.01945	7.07×

[cursor]

Stale Bugbot comment from a previous run.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit ea0c4fb. Configure here.}