docs(tests): no_eof growing-file notes say a mid-record append raises ValueError (#1265) #2328

Is your feature request related to a problem? Please describe.
The module docstring of tests/foundation/test_extraction_no_eof.py ("Worth knowing while reading that ...") and the comment in NoEOFOverAGrowingFileStopsAtWhatIsThere.test_a_file_appended_to_after_the_read_yields_only_what_was_there say that an append landing mid-record raises ValueError: read length must be non-negative or -1 on every tree, and call it not addressed. That is no longer true. Driving examples/captures/in.pcap through every split point from 25 to 604, under no no_eof, no_eof with no append, an append at the first end of stream, and an append at the second, raises nothing on main. The same sweep raises that ValueError on nearly every split at 79494a089 (#639). The fix was bda27e073 (#1331, for #1265): the ValueError is present at its parent and gone at that commit.

Describe the solution you'd like
Replace both notes with what happens now. A record cut inside its header ends extraction at the last whole record. A record cut inside its data is yielded with the octets present. The test's split = 430 can stay, because a boundary is still what makes the five-frame assertion exact.

Describe alternatives you've considered
Pinning the mid-record behaviour with its own case. That is a test, not a prose fix, so it would be a separate change.

Additional context
No open issue tracks this. #1583 is the earlier timer race in the same test.