fix: save rss changelogs to input data

[hassayag]

The new fetching process loads input-data/changelogs files first. Any output data is then also written to input data, as input-data is to be treated as a persistent storage - this is handled by the deploy script, in which it commits any input-data changes.

Since the RSS feed can only see recent changelog data, the rest will now be made up by this input-data. Therefore output-data can be consistently reproduced from nothing. Because of that consistency, deadlock-data has been reverted back to the simple data directory, which reflects what output-data would look like locally

Parsed data in deadlock-data PR

[Surxe]

I still have roughly 2 hours of available effort a day due to sickness, so I skimmed very lightly, but this looks very good! Like the idea, even if it means input-data may have slightly more diffs, I think its well worth. Also I like the simplified external calls for changelog fetcher