perf(container): cap the dependency chain backtrace

[osbre]

resolveChain() previously captured the entire call stack on every container resolution but only needed the immediate caller's frame.

Added a test covering the reported origin file, which lacked coverage previously.

Impact

Micro-optimization territory 😄

• Container lookups: Changes performance from linear to constant relative to stack depth, dropping from 3.23 µs down to 1.43 µs at depth 60.
• App performance: Saves about 35 µs per request on a trivial route. On an application doing real-world work it's a tiny fraction of a percent, but it's a tidy way to remove unnecessary overhead.

[github-actions]

Benchmark Results

Comparison of perf/container-backtrace-limit against 3.x (cf035264805564eb789f2741734e481aa3696b4d).

Open to see the benchmark results

Benchmark	Set	Mem. Peak	Time	Variability
QueryExecutionBench(benchExecuteInsert)	-	4.400mb 0.00%	9.120μs -6.78%	±1.78% +31.69%
GenericRouterBench(benchDispatch)	Static route	5.264mb 0.00%	40.799μs -7.76%	±0.47% -9.36%
GenericRouterBench(benchDispatch)	Dynamic route	5.268mb 0.00%	48.127μs -7.37%	±0.50% -1.98%
GenericRouterBench(benchDispatch)	Dynamic short	5.268mb 0.00%	48.358μs -6.28%	±0.72% +162.19%
GenericRouterBench(benchDispatchWithoutMiddleware)	Static route	5.263mb 0.00%	33.820μs -5.82%	±0.46% +5.33%
ViewRenderBench(benchPlainHtml)	-	22.035mb 0.00%	421.681μs -7.79%	±1.79% -33.05%
ViewRenderBench(benchControlFlow)	-	44.501mb 0.00%	639.534μs -5.41%	±1.71% +55.27%
DiscoveryScanBench(benchFullDiscoveryScan)	-	25.857mb 0.00%	20.479ms +7.16%	±1.16% +9.58%

_{Generated by phpbench against commit f39b24c}