[flthibaud]

PostgreSQL refuses a NUL byte or invalid UTF-8 in a text column. A feed item carrying one (a JSON Feed title with \u0000 gets through gofeed) failed its insert on every sync. A failed item keeps the feed's ETag from advancing, so the whole feed was downloaded again and the page scraped again every 15 minutes, forever.

• New cleanArticleString applied to title, description, author, HTML and Markdown in createArticleFromItem.
• New test TestCreateArticleFromItemCleansUnstorableStrings.

Tested with gofmt, go vet ./... and go test -race ./... against a migrated postgres:16 (no test skipped).

Closes #42

[flthibaud]

Review: OK to merge.

• Code and test by Mistral. The test did not compile (gofeed.Author does not exist, the type is gofeed.Person) and a comment had a garbled escape; I fixed both.
• The article URL and image URL are not cleaned. They come from the same feed fields, but a NUL there would already fail url.Parse or the HTTP fetch upstream for the link; the image URL is a residual gap.
• PostgreSQL's refusal of a NUL byte was confirmed on a real database earlier (invalid byte sequence for encoding "UTF8": 0x00).