Fix diff generation for lines without newline

[ad1217]

Rather than keeping the line endings before passing to diff, then joining the resulting lines with '', strip the line endings and join
with '\n'. This ensures a that each output line has exactly one trailing newline.

Possible issues:

• This will prevent detection of a difference in line endings.
• Joining with os.linesep might be better for Windows, but only if it is going to a text file, and I think most Windows stuff handles lines terminated with just LF fine now.

[polyzen]

Before:

@@ -1,3 +1,3 @@
-1.5.8
-2020-07-01T04:51:11
-707ed1bdffd6b8bec2d74dca36cc2665081c8db8590241bf6a72a3378928dd65+1.5.9
+2020-07-23T04:58:02
+6f3e18befd0f26118c6c6ddf39e2ef18366ff9501cbcbf664ffde669e3c2d132

After:

@@ -1,3 +1,3 @@
-1.5.8
-2020-07-01T04:51:11
-707ed1bdffd6b8bec2d74dca36cc2665081c8db8590241bf6a72a3378928dd65
+1.5.9
+2020-07-23T04:58:02
+6f3e18befd0f26118c6c6ddf39e2ef18366ff9501cbcbf664ffde669e3c2d132

With this filter:
https://github.com/polyzen/dotfiles/blob/3b83c549d2e497098564ad7fdbb44c5050379231/misc/.config/urlwatch/hooks.py#L52

[thp]

This is a good change, I've ran into the issue as described by @polyzen already a few times, good that this is now fixed. I think most people don't care about newline changes, and if they do, they can always use an external diff_tool or the hexdump filter.

Change documented here: c3b5c24