[jay23606]

Bug

Geometry.AsBinary and the per-type AsBinary methods called AppendWKB(nil). A nil slice has zero capacity, so the output buffer was regrown several times while the WKB was written.

Change

• Each geometry type gets a bytesInWKB() method that computes the exact WKB size. AsBinary allocates make([]byte, 0, g.bytesInWKB()) and appends into it, so the buffer is allocated once.
• The regression test checks that cap equals len for each geometry type and that AsBinary makes exactly one allocation. A second test checks that bytesInWKB matches len(AsBinary()).
• BenchmarkMarshalWKB now calls b.ReportAllocs(), so the allocation count is reported alongside timing.

Testing

• go test ./geom/ -run TestBytesInWKBMatchesAsBinary -count=1 passes on the fix. On the base commit it fails (go -C .pr-go/base test ./geom/ -run TestBytesInWKBMatchesAsBinary -count=1).
• go test ./geom/ -run TestBytesInWKB -count=1 -v passes on the fix, covering both the length check and the allocation check.
• go vet ./geom/ is clean.
• go test ./geom/ -run ^$ -bench BenchmarkMarshalWKB/polygon -benchtime=200x (one run each, so the timings are noisy): on the fix, the n=10000 polygon is 34,124 ns/op with 1 allocs/op. On base, it is 85,596 ns/op.
• go test ./geom/ -count=1 has one failure, TestGeoJSONSyntaxError.

Regression test: fails on base (go -C .pr-go/base test ./geom/ -run TestBytesInWKBMatchesAsBinary -count=1), passes on fix (go test ./geom/ -run TestBytesInWKBMatchesAsBinary -count=1)

Fixes #368

[peterstace]

Thanks for the PR! I've got some requests for change.