Rugged::Diff::Delta #new_file / #old_file method got a wrong encoded [:path] string

when I call new_file[:path] / old_file[:path] of Rugged::Diff::Delta, I got result as follow:

  [{:path=>"test/.keep", :type=>"added"},
   {:path=>"test/\xE4\xB8\xAD\xE6\x96\x87\xE5\x90\x8D\xE6\xB5\x8B\xE8\xAF\x95\xE4\xB8\x80.txt", :type=>"added"},
   {:path=>"test/\xE4\xB8\xAD\xE6\x96\x87\xE5\x90\x8D\xE6\xB5\x8B\xE8\xAF\x95\xE4\xB8\x89.txt", :type=>"added"},
   {:path=>"test/\xE4\xB8\xAD\xE6\x96\x87\xE5\x90\x8D\xE6\xB5\x8B\xE8\xAF\x95\xE4\xBA\x8C.txt", :type=>"added"},
   {:path=>"\xE4\xB8\xAD\xE6\x96\x87.  \xE5\x91\xBD\xE5\x90\x8D1.txt", :type=>"added"}]}

when I copy the path value to my console pry/irb, I got the right string

It seems to be an incorrectly decoded problem

[66] pry(main)> "test/\xE4\xB8\xAD\xE6\x96\x87\xE5\x90\x8D\xE6\xB5\x8B\xE8\xAF\x95\xE4\xB8\x80.txt"
=> "test/中文名测试一.txt"
[67] pry(main)> "test/\xE4\xB8\xAD\xE6\x96\x87\xE5\x90\x8D\xE6\xB5\x8B\xE8\xAF\x95\xE4\xB8\x89.txt"
=> "test/中文名测试三.txt"
[68] pry(main)> "test/\xE4\xB8\xAD\xE6\x96\x87\xE5\x90\x8D\xE6\xB5\x8B\xE8\xAF\x95\xE4\xBA\x8C.txt"
=> "test/中文名测试二.txt"

Reproduce the issue through Rugged::Diff::Delta#new_file and #old_file using the non-ASCII paths shown. Trace how path strings are encoded in the Rugged Ruby bindings and libgit2 boundary; done means these entry points return the expected Unicode paths and a regression test covers them.