feat: add escape analysis MVP

[MeteorsLiu]

This draft starts the Phase 1 LLVM-IR escape-analysis MVP proposed in #2244.

The implementation includes:

• Add a sparse per-allocation LLVM use walker with conservative operand-role handling and exact forwarding through proven local alloca slots.
• Cache pointer-parameter summaries for heap, mutator, callee, result, and alignment facts, including direct-call composition and recursive fixed-point evaluation.
• Rewrite bounded local AllocZ and AllocU calls before large-aggregate and C ABI lowering, preserving zero initialization and rejecting loops, unsupported sizes, alignments, and address spaces.
• Drive focused pass coverage from literal LITTEST fixtures where in.txt contains input LLVM IR and out.txt contains the complete transformed module, plus a build-pipeline integration fixture.
• Keep call-argument alignment propagation and recursive alignment-level saturation as explicit follow-up correctness work before this draft is ready for review.

This establishes the LLGo-owned analysis and rewrite boundary while keeping unknown calls and unsupported LLVM behavior conservative.

[github-actions]

LLGo baseline benchmarks

97ae05374b21 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs main	Build	vs main	Run	vs main
Linux	cprintf	18544 B	new	348.361 ms	new	1.301 ms	new
Linux	fmtprintf	2214984 B	new	3.399 s	new	2.498 ms	new
Linux	println	71504 B	new	347.496 ms	new	1.635 ms	new
macOS	cprintf	84672 B	new	440.807 ms	new	5.018 ms	new
macOS	fmtprintf	2360880 B	new	3.958 s	new	20.668 ms	new
macOS	println	125712 B	new	426.287 ms	new	4.534 ms	new

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs main
Linux	BenchmarkLookupPCRandom	13.210 ns/op	new
Linux	BenchmarkMergeCompilerFlags	150.600 ns/op	new
Linux	BenchmarkMergeLinkerFlags	94.250 ns/op	new
Linux	BenchmarkChannelBuffered	35.270 ns/op	new
Linux	BenchmarkChannelHandoff	31298 ns/op	new
Linux	BenchmarkDefer	45.400 ns/op	new
Linux	BenchmarkDirectCall	1.557 ns/op	new
Linux	BenchmarkGlobalRead	1.868 ns/op	new
Linux	BenchmarkGlobalWrite	2.488 ns/op	new
Linux	BenchmarkGoroutine	35489 ns/op	new
Linux	BenchmarkInterfaceCall	8.099 ns/op	new
Linux	BenchmarkRuntimeGetG	2.510 ns/op	new
macOS	BenchmarkLookupPCRandom	12 ns/op	new
macOS	BenchmarkMergeCompilerFlags	113.300 ns/op	new
macOS	BenchmarkMergeLinkerFlags	74.890 ns/op	new
macOS	BenchmarkChannelBuffered	20.890 ns/op	new
macOS	BenchmarkChannelHandoff	6718 ns/op	new
macOS	BenchmarkDefer	27.610 ns/op	new
macOS	BenchmarkDirectCall	0.971 ns/op	new
macOS	BenchmarkGlobalRead	0.958 ns/op	new
macOS	BenchmarkGlobalWrite	0.943 ns/op	new
macOS	BenchmarkGoroutine	37800 ns/op	new
macOS	BenchmarkInterfaceCall	4.241 ns/op	new
macOS	BenchmarkRuntimeGetG	1.969 ns/op	new

No main baseline exists yet; all metrics are marked new.

Warning

• The rendered benchmark data could not be pushed.

[codecov]

Codecov Report

❌ Patch coverage is 90.44944% with 34 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/escape/escape.go	90.39%	23 Missing and 9 partials ⚠️
internal/littest/littest.go	90.47%	1 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!