Fix processing of attachments encoded as Latin 1

When body has been encoded at Latin 1 / ISO-8859-1 we should ensure the
strings are converted back to UTF-8 before generating their hexdigest.

This fixes applying masks and censor rules to some attachments.

Fixes #7915

lib/mail_handler/backends/mail_backend.rb

@@ -404,7 +404,7 @@ def attempt_to_find_original_attachment_attributes(mail, body:, nested: false)
         def calculate_hexdigest(body)
           # ensure bodies have the same line endings
           Digest::MD5.hexdigest(Mail::Utilities.binary_unsafe_to_lf(
-            body.rstrip
+            convert_string_to_utf8(body.rstrip).string
           ))
         end
 

spec/lib/mail_handler/backends/mail_backend_spec.rb

@@ -461,6 +461,7 @@
         add_file filename: 'lf.txt', content: "bar\nbar"
         add_file filename: 'crlf-non-ascii.txt', content: "Aberdâr\r\n"
         add_file filename: 'lf-non-ascii.txt', content: "Aberdâr\n"
+        add_file filename: 'latin1.txt', content: "naïve"
         add_file filename: 'mail.eml', content: mail_attachment
         add_file filename: 'uuencoded.eml', content: mail_with_uuencoded
       end
@@ -492,6 +493,11 @@
       it { is_expected.to include(body: "Aberdâr\n") }
     end
 
+    context 'when binary body encoded as Latin 1 / ISO-8859-1' do
+      let(:body) { "na\xEFve".b }
+      it { is_expected.to include(body: "naïve") }
+    end
+
     context 'when attached email headers are different' do
       let(:body) do
         <<~EML