fix(application): an HTTP/2 SETTINGS frame cut short by the capture raises instead of parsing #2659

An HTTP/2 SETTINGS frame cut short by the capture raises a loud FieldValueError from ListField.unpack's infinite-loop guard (#431) instead of parsing as far as it goes. Even if it parsed, it could not rebuild as captured, because HTTPv2.make recomputes the frame Length from the settings it holds. The #2650 worker found this while comparing a change to the shared ListField, which it reverted. It has not been reproduced independently.

Expected: a SETTINGS frame cut inside its settings parses the settings present, records the short read, and rebuilds the captured octets. That matches how #2650 handles IGMPv3 Reports and how SCTP's ChunkListField sizes its list.

To do:

• Reproduce on main.
• Give the settings list a field sized by the octets present, or reuse fix(internet): a truncated IGMPv3 Report falls back to Raw instead of parsing what it carries #2650's pattern once it merges.
• Pass the captured Length through make() the way fix(internet): a truncated IGMPv3 Report falls back to Raw instead of parsing what it carries #2650 passes num_group_records.
• Test cuts at every offset, plus a whole frame as control.