Preserve ordering of comments in .po files

An option to preserve the order of comment fields in the translation data would be very useful for me. This library almost works for my use case, but parsing a .po file with the library, modifying the translation data, and re-compiling the translation data back into a .po file with the library changes the ordering of comment fields. For example:

-#. ~ Snippet in category "tip"
 #: data/core/tips.json
+#. ~ Snippet in category "tip"
 msgid "Moose may not be your friend."
 msgstr "Elche sind vielleicht nicht deine Freunde."

This seems consistent behavior, #. lines are always moved below #: lines. This causes diffs for translation files to become very large and hard to review. This is inconvenient for our workflow. The original translation files are generated with a process that orders the comments like this, which I can't affect.

Trace the library's .po parsing and compilation paths, focusing on how comment fields are stored and emitted. Done means parsing, modifying, and recompiling a .po file preserves the original relative order of comment fields such as #. and #:, avoiding unrelated translation-file diffs.