…into develop-rss

The previous test compared the xml as a string to an expected string. This worked occasionally because the xml writer will sometimes place attributes in a random order.

Also a bunch of formatting changes while I tried to eliminate warnings

…into develop-rss

Change import from ET to ElemTree and change whitespace location in expected string to satisfy Flake8. This doesn't affect the test, but unclogs the test output a little.

Recursive test fails on assertion rather than returning bool to bubble up through the nested generators.

Strip newlines from expected string before parsing in ElementTree

Strips all non-ascii characters when writing the xml document.

Writes a new xml document if the existing document is empty or just whitespace

Some episode descriptions contain non-ascii characters, and this can cause errors. This commit simply ignores those characters, which is safe because hashes and urls must be ascii by definition.

qBittorrent (and possibly others) do not support xml with ascii encoding, so this commit forces utf-8/unicode in ElementTree and when reading/writing.