[Superseded by #2327] crosscompile: support cache file locks on Windows

[cpunion]

Superseded by #2327.

The Windows cache-file locking change from this PR has been replayed as its own commit on the latest main in #2327, together with the related target identity, metadata mapping, and native Windows smoke coverage.

This PR remains Draft only as historical review context. New Windows work should depend on #2327 rather than this PR.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 11 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/crosscompile/fetch.go	0.00%	7 Missing ⚠️
internal/crosscompile/filelock_unix.go	0.00%	4 Missing ⚠️

📢 Thoughts on this report? Let us know!

[github-actions]

LLGo baseline benchmarks

3f4767f11d69 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs base	Build	vs base	Run	vs base
Linux	cprintf	18656 B	+0.0%	284.096 ms	-2.9% (better)	1.272 ms	+2.0% (worse)
Linux	fmtprintf	1874264 B	+0.0%	2.711 s	+3.6% (worse)	3.330 ms	+1.6% (worse)
Linux	println	67992 B	+0.0%	279.385 ms	-7.6% (better)	1.600 ms	+2.4% (worse)
macOS	cprintf	84624 B	+0.0%	302.179 ms	+2.2% (worse)	2.649 ms	+5.4% (worse)
macOS	fmtprintf	1889968 B	+0.0%	2.344 s	+6.7% (worse)	14.291 ms	+12.2% (worse)
macOS	println	121168 B	+0.0%	297.058 ms	-1.8% (better)	3.154 ms	-7.9% (better)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs base
Linux	BenchmarkLookupPCRandom	13.260 ns/op	-0.2% (better)
Linux	BenchmarkMergeCompilerFlags	150.400 ns/op	-0.1% (better)
Linux	BenchmarkMergeLinkerFlags	94.200 ns/op	-0.2% (better)
Linux	BenchmarkChannelBuffered	35.040 ns/op	-0.1% (better)
Linux	BenchmarkChannelHandoff	27715 ns/op	+6.9% (worse)
Linux	BenchmarkDefer	45.110 ns/op	+0.3% (worse)
Linux	BenchmarkDirectCall	1.558 ns/op	+0.1% (worse)
Linux	BenchmarkGlobalRead	1.556 ns/op	-0.1% (better)
Linux	BenchmarkGlobalWrite	2.486 ns/op	-0.0% (better)
Linux	BenchmarkGoroutine	42151 ns/op	+36.5% (worse)
Linux	BenchmarkInterfaceCall	7.783 ns/op	-0.1% (better)
Linux	BenchmarkRuntimeGetG	2.492 ns/op	-0.4% (better)
macOS	BenchmarkLookupPCRandom	12.660 ns/op	+6.5% (worse)
macOS	BenchmarkMergeCompilerFlags	102.800 ns/op	-3.2% (better)
macOS	BenchmarkMergeLinkerFlags	80.010 ns/op	+23.4% (worse)
macOS	BenchmarkChannelBuffered	22.780 ns/op	-29.8% (better)
macOS	BenchmarkChannelHandoff	8492 ns/op	-18.4% (better)
macOS	BenchmarkDefer	28.320 ns/op	-27.8% (better)
macOS	BenchmarkDirectCall	1.029 ns/op	-9.2% (better)
macOS	BenchmarkGlobalRead	1.277 ns/op	+27.3% (worse)
macOS	BenchmarkGlobalWrite	1.047 ns/op	+5.8% (worse)
macOS	BenchmarkGoroutine	28825 ns/op	+16.3% (worse)
macOS	BenchmarkInterfaceCall	4.514 ns/op	-14.7% (better)
macOS	BenchmarkRuntimeGetG	2.080 ns/op	-11.9% (better)

Compared with f65ddd1b0b4e measured in the same runner job.