Consider using SafeHandle in Libgit2Object instead of void*

We've had a couple of bug reports against dotnet/runtime with rarely-reproduceable crashes coming from libgit2sharp. We believe that one possible reason could be an interop anti-pattern in Libgit2Object - void*/IntPtr representation of a native handle that can be freed in finalizer. E.g. in this case the finalizer in Libgit2Object may end up calling native free here (and other overloads).

The reason why it's called an anti-pattern can be explained by a short repro in this issue: https://github.com/dotnet/runtime/issues/103522 and a general solution is to use SafeHandle for such handles. Also, see https://learn.microsoft.com/en-us/dotnet/standard/native-interop/best-practices

Reproduction steps

Expected behavior

Actual behavior

Version of LibGit2Sharp (release number or SHA1)

Operating system(s) tested; .NET runtime tested

LibGit2Sharp/Core/Handles/Libgit2Object.cs の 88 行目付近から始め、続いて LibGit2Sharp/Core/Handles/Objects.cs の 502-505 行目付近にある finalizer と native-free overloads を追跡します。これらの所有権パスを SafeHandle のガイダンスおよびリンクされた runtime レポートと比較します。完了条件は、関連する native handle のライフタイムパスが安全に表現され、報告された finalizer race なしに解放されることです。