PR #1094 generated an HTML description for any episode that lacked one.
That however increased the database size (almost double in worst case)
because it was storing both text and html versions of each description.
This fixes that by storing the episode thumbnail URL in the database and
generating the HTML description only when shownotes are drawn.
The text description is now cleared for episodes with an HTML
description. This further reduces database size for feeds that provide
both. It also fixes an issue for feeds that provide different text and
HTML descriptions, because the short description would show the text
description and shownotes would show the HTML description. And EQL only
searched the text descriptions, which might not match what the user sees
in the shownotes.
Implement a faster method for acquiring the
episode object from a download URL or filename.
This improves performance for episode status
updates and episode metadata requests over D-Bus.