perf: ~2.9x faster request dispatch, and add a benchmark suite

[gmpassos]

Adds benchmark/ — an in-process suite for the request path — and fixes the costs it exposed.

Benchmark	Before	After
APIRouteHandler.call (direct)	2.128 µs	0.524 µs	4.1×
APIRoot.call: ping	2.758 µs	0.901 µs	3.1×
APIRoot.call: echo (parameters)	3.057 µs	1.203 µs	2.5×
APIRoot.call: json (entity payload)	3.081 µs	1.222 µs	2.5×

Measured back-to-back on one machine (dart run benchmark/bones_api_benchmark.dart). In-process — no socket, no HTTP client — so this is framework overhead only.

Findings, in order of impact

1. Log messages were formatted even when nothing consumed them.

LoggerHandler._logRootMsg called _buildMsg on every record before deciding whether any destination wanted it. _buildMsg renders the timestamp, pads and truncates the isolate and logger names, and looks the current APIRequest up in the record's Zone. The result was then discarded whenever no destination (logAllTo/logErrorTo/logDbTo/console) was configured — which is the default. That is ~1 µs per record, and a route call emits two (CALL> and RESPONSE>).

Now guarded by _hasLogDestination, which is deliberately conservative: it may answer true and let the existing dispatch decide, but never false while a destination exists.

2. APIRouteHandler re-interpolated its CALL> message per request — including stringifying the declared parameters Map — although module, routeName and parameters are fixed once a route is registered. Now built once, rebuilt only if parameters is replaced. Same for the RESPONSE> prefix.

3. APIRoot._callImpl read pathParts[0], and pathParts copies the backing list on every access. Uses pathPartFirst.

4. APIServer.toAPIRequest copied the query-parameters Map twice — it had just built it.

Being straight about magnitudes: #1 is essentially the whole win. #3 and #4 measured ~2% together, at the edge of run-to-run noise on this machine; I kept them because they are strictly less work, not because they move the number. #2 helps most on routes that declare parameters (the benchmark routes declare none, so it barely shows there).

API addition

The routes builder now forwards config: on any/get/post/put/delete/patch/head, matching APIModule.addRoute:

routes.get('ping', handler, config: const APIRouteConfig(log: false));

An APIRouteConfig was previously only reachable through addRoute, so per-route logging could not be turned off through the usual API. That matters: route logging still costs roughly 4× the rest of a trivial dispatch even after this PR (0.901 µs with logging vs 0.717 µs without). The suite measures both paths side by side so the trade-off stays visible.

Benchmarks

dart run benchmark/bones_api_benchmark.dart
dart run benchmark/bones_api_benchmark.dart --emit-baseline   # to compare a change

Layered by design — request construction, path accessors, routing, handler, module, root — so a regression can be attributed rather than just observed at the top. Dependency-free on purpose: benchmark/ ships inside the published package, so a benchmarking dependency would land on every consumer. _baseline is intentionally left empty, since throughput is hardware-specific; see benchmark/README.md.

Testing

bones_api_logging_test.dart had no coverage of destination routing — exactly what change #1 touches. Added four tests for it, and verified they are real: forcing the new guard to drop everything makes 3 of them fail.

Gate	Result
dart test --exclude-tags docker	783 passed
dart analyze --fatal-infos --fatal-warnings .	clean
dart format -o none --set-exit-if-changed .	clean
dart pub publish --dry-run	0 warnings

Docker-tagged suites not run locally (no daemon).

Version bumped to 1.15.0 with a CHANGELOG entry — drop that commit hunk if you'd rather release separately.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 68.15%. Comparing base (ed1a703) to head (d30685d).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #150      +/-   ##
==========================================
+ Coverage   68.13%   68.15%   +0.02%     
==========================================
  Files          66       66              
  Lines       22114    22129      +15     
==========================================
+ Hits        15067    15083      +16     
+ Misses       7047     7046       -1     

Flag	Coverage Δ
unittests	68.15% <100.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.