incorrect assumption?

hello - thanks for the repo. However, thought I'd share something I noticed with you or anyone else reading over a decade later :). You're allocating memory into the Go heap [here](https://github.com/felixge/go-cgo-finalizer/blob/master/finalizer.go#L56). Thus to me it seems you're creating _go_ heap memory pressure, and this test program is not demonstrating that:

> yes, runtime.SetFinalizer() works well for automatically free()'ing cgo memory allocations!

In a real scenario, the C code would be allocating into the C heap. The Go GC only cares about the Go heap, so it's not going to see any memory pressure, GC will not run, and thus neither will the finalizers. Here's a [more realistic test script](https://gist.github.com/timruffles/48bfa3bd5a625d80e7c73d0eacfc8335) using `malloc` which demonstrates that you cannot rely on `runtime.SetFinalizer` to keep CGO memory usage within any defined bound.