fix(rss): detect empty or HTML responses that feedparser treats as non-bozo

[2233admin]

Problem

The W1 schema-drift wiring in #35 handles RSS parse failures when feedparser reports bozo=True. A different blind spot remains: HTTP 200 responses containing an HTML error page or an empty body can be parsed as bozo=False with entries=[].

Those responses currently look like a successful empty feed, so no structured error_type reaches the control recorder and the SCHEMA_DRIFT sensor does not fire.

Reproduction

Observed with feedparser using both:

• an HTML error page body
• an empty response body

In both cases the result can be bozo=False and entries=[].

Expected behavior

Classify suspicious empty/non-feed responses explicitly without treating legitimate empty feeds as schema drift.

Acceptance criteria

• define a conservative distinction between a valid empty feed and a non-feed/empty HTTP response
• return a structured error type for the suspicious cases
• ensure that error type maps to the intended control error kind
• add unit tests for HTML error pages, empty bodies, and a valid feed with zero entries
• document any content-type or body-shape heuristics used

Scope note

Keep this separate from #35. That PR repairs error propagation for known parser failures; this issue covers response-quality detection before/around parser classification.