[Based on #126] debug: add schema-driven GDB runtime support

[cpunion]

Part of xgo-dev#2164 Task 4. Based on #126.

LLGo debug artifacts could be transported to GDB, but GDB still exposed raw runtime layouts and section GC could discard the debugger ABI marker on embedded links.

This change:

• adds an embedded, versioned-schema GDB adapter and launcher with raw-GDB fallback;
• presents strings, slices, interfaces, funcs, maps, channels, and named values;
• exposes goroutine listing and thread-backed goroutine commands with M/P ownership checks;
• retains and aligns the native debugger ABI record through final ELF/Mach-O links;
• exercises native Linux GDB and Cortex-M QEMU GDB Remote in CI while preserving the existing LLDB remote check.

Local validation:

• macOS arm64: changed-package tests passed; aggregate coverage 78.1% (cmd/internal/gdb 78.2%, internal/debuginfo 100%, internal/build 77.7%).
• Linux arm64: native GDB integration passed; adapter/debuginfo aggregate coverage 92.1%; internal/build coverage 77.0%.
• Linux arm64 + Cortex-M QEMU: stripped load-image equivalence, DWARF verification, GDB Remote runtime presentation, and LLDB Remote source/locals/backtrace all passed.

[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

6f2fd2a399b0 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs main	Build	vs main	Run	vs main
Linux	cprintf	19344 B	+4.3% (worse)	402.131 ms	-2.5% (better)	1.377 ms	-11.3% (better)
Linux	fmtprintf	2218824 B	+0.2% (worse)	3.381 s	-12.3% (better)	2.736 ms	-16.8% (better)
Linux	println	73832 B	+1.6% (worse)	382.443 ms	-4.4% (better)	1.733 ms	+3.6% (worse)
macOS	cprintf	84672 B	+0.0%	324.509 ms	-21.7% (better)	2.757 ms	-23.2% (better)
macOS	fmtprintf	2361696 B	+0.7% (worse)	2.703 s	-31.4% (better)	18.705 ms	-20.0% (better)
macOS	println	125888 B	-0.9% (better)	326.006 ms	-21.4% (better)	3.692 ms	-49.1% (better)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs main
Linux	BenchmarkLookupPCRandom	12.240 ns/op	-9.2% (better)
Linux	BenchmarkMergeCompilerFlags	144.200 ns/op	-6.7% (better)
Linux	BenchmarkMergeLinkerFlags	93.660 ns/op	-7.5% (better)
Linux	BenchmarkChannelBuffered	36.780 ns/op	+5.8% (worse)
Linux	BenchmarkChannelHandoff	24655 ns/op	-23.0% (better)
Linux	BenchmarkDefer	45.380 ns/op	-19.4% (better)
Linux	BenchmarkDirectCall	1.760 ns/op	+12.6% (worse)
Linux	BenchmarkGlobalRead	1.757 ns/op	+12.6% (worse)
Linux	BenchmarkGlobalWrite	2.808 ns/op	+12.6% (worse)
Linux	BenchmarkGoroutine	29810 ns/op	-60.6% (better)
Linux	BenchmarkInterfaceCall	8.797 ns/op	+12.9% (worse)
Linux	BenchmarkRuntimeGetG	2.110 ns/op	-60.2% (better)
macOS	BenchmarkLookupPCRandom	13.710 ns/op	+10.2% (worse)
macOS	BenchmarkMergeCompilerFlags	125.300 ns/op	+0.6% (worse)
macOS	BenchmarkMergeLinkerFlags	77.400 ns/op	+11.6% (worse)
macOS	BenchmarkChannelBuffered	22.280 ns/op	-19.7% (better)
macOS	BenchmarkChannelHandoff	9216 ns/op	+31.5% (worse)
macOS	BenchmarkDefer	28.760 ns/op	-23.2% (better)
macOS	BenchmarkDirectCall	1.017 ns/op	-30.0% (better)
macOS	BenchmarkGlobalRead	1.027 ns/op	-15.2% (better)
macOS	BenchmarkGlobalWrite	1.018 ns/op	-31.5% (better)
macOS	BenchmarkGoroutine	29357 ns/op	-7.3% (better)
macOS	BenchmarkInterfaceCall	4.432 ns/op	-17.9% (better)
macOS	BenchmarkRuntimeGetG	2.313 ns/op	-10.2% (better)

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

Warning

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