[Based on #101] debug: add LLDB goroutine views

[cpunion]

Summary

• add llgo goroutines, backed by a versioned live-G list read by LLDB without executing inferior code
• report each live goroutine state, parent ID, and G/M/P ownership
• keep O(1) registration/removal correct for runtime-owned and lazily-created G lifecycles
• read every goroutine field and status name from the common debugger schema
• fix G/M/P ID allocation for the old-value semantics of LLGo atomic fetch-add

Advances xgo-dev#2164 and xgo-dev#2154. Goroutine backtraces remain in the dependent #106. Based on #101.

Tests

• macOS arm64, Apple LLDB 1703: 275/275
• macOS and Linux amd64: runtime G/M/P and live-G lifecycle tests
• macOS and Linux amd64: go test ./cmd/internal/lldb ./cmd/llgo
• cmd/internal/lldb coverage: 87.5%
• Python syntax check and git diff --check

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[github-actions]

LLGo baseline benchmarks

53b1e7c82063 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs main	Build	vs main	Run	vs main
Linux	cprintf	18640 B	+0.5% (worse)	345.491 ms	-16.2% (better)	1.299 ms	-16.3% (better)
Linux	fmtprintf	2218464 B	+0.2% (worse)	3.345 s	-13.2% (better)	2.453 ms	-25.4% (better)
Linux	println	72088 B	-0.8% (better)	352.633 ms	-11.9% (better)	1.632 ms	-2.5% (better)
macOS	cprintf	84672 B	+0.0%	304.068 ms	-26.6% (better)	2.281 ms	-36.4% (better)
macOS	fmtprintf	2361632 B	+0.7% (worse)	2.599 s	-34.0% (better)	17.959 ms	-23.2% (better)
macOS	println	125824 B	-1.0% (better)	293.627 ms	-29.2% (better)	3.182 ms	-56.2% (better)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs main
Linux	BenchmarkLookupPCRandom	13.230 ns/op	-1.9% (better)
Linux	BenchmarkMergeCompilerFlags	154.600 ns/op	+0.1% (worse)
Linux	BenchmarkMergeLinkerFlags	99.730 ns/op	-1.5% (better)
Linux	BenchmarkChannelBuffered	35.210 ns/op	+1.3% (worse)
Linux	BenchmarkChannelHandoff	27993 ns/op	-12.6% (better)
Linux	BenchmarkDefer	43.630 ns/op	-22.5% (better)
Linux	BenchmarkDirectCall	1.556 ns/op	-0.4% (better)
Linux	BenchmarkGlobalRead	1.556 ns/op	-0.3% (better)
Linux	BenchmarkGlobalWrite	2.481 ns/op	-0.5% (better)
Linux	BenchmarkGoroutine	31879 ns/op	-57.8% (better)
Linux	BenchmarkInterfaceCall	8.096 ns/op	+3.9% (worse)
Linux	BenchmarkRuntimeGetG	2.177 ns/op	-59.0% (better)
macOS	BenchmarkLookupPCRandom	11.220 ns/op	-9.8% (better)
macOS	BenchmarkMergeCompilerFlags	127.500 ns/op	+2.4% (worse)
macOS	BenchmarkMergeLinkerFlags	74.920 ns/op	+8.0% (worse)
macOS	BenchmarkChannelBuffered	20.550 ns/op	-25.9% (better)
macOS	BenchmarkChannelHandoff	6530 ns/op	-6.8% (better)
macOS	BenchmarkDefer	24.940 ns/op	-33.4% (better)
macOS	BenchmarkDirectCall	0.945 ns/op	-35.0% (better)
macOS	BenchmarkGlobalRead	0.988 ns/op	-18.4% (better)
macOS	BenchmarkGlobalWrite	1.035 ns/op	-30.4% (better)
macOS	BenchmarkGoroutine	22325 ns/op	-29.5% (better)
macOS	BenchmarkInterfaceCall	4.081 ns/op	-24.4% (better)
macOS	BenchmarkRuntimeGetG	1.889 ns/op	-26.7% (better)

Compared only with the latest matching platform in the main series.

Warning

• Persistent publishing is unavailable because no data token is configured.