docs(changelog): state the 1.2.0 regression and CWE-22 affected range

[dani1005]

Summary

The rewritten 1.1.4 section now documents the code shipped as everos==1.1.4
on PyPI, which correctly lists the knowledge-upload path traversal (CWE-22) as
fixed. Read top-down, however, that leaves a 1.2.0 reader concluding the fix
predates their version and they are therefore safe — the opposite of the truth.
1.2.0 was built from a branch that never received those fixes, so it regressed
to pre-1.1.4 behaviour, and the changelog says so nowhere.

Three changes, no history rewritten:

1. The 1.2.1 Fixed entry now names what 1.2.0 regressed, instead of
   describing the merge as internal branch hygiene ("backfilled into main …
   so subsequent releases build on top of them"), which tells a user nothing
   about their own exposure.
2. A Security section under 1.2.1 carries the affected-version range. The
   range is discontinuous — everything before 1.1.4, plus 1.2.0; 1.1.4
   itself is not affected — so it cannot be inferred from "fixed in 1.2.1".
3. A one-line note under 1.2.0 points forward, covering the reader who looks up
   the version they are running rather than the newest one.

This brings CHANGELOG.md in line with the published
v1.2.1 release notes,
which already carry the same Security block.

Area

• Architecture method
• Benchmark
• Use case
• Documentation
• Developer experience
• CI, build, or release

Verification

Affected-version range verified against the published artifacts, not inferred —
each sdist downloaded from PyPI and _write_original_file inspected:

  everos 1.1.4  -> _safe_original_filename + resolve().is_relative_to  present
  everos 1.2.0  -> target = original_dir / source_name, no validation  ABSENT
  everos 1.2.1  -> containment present

python3 scripts/check_commit_messages.py HEAD^..HEAD  -> pass
No line in the diff exceeds 80 columns (matches surrounding style).
Documentation-only change; no code or tests affected.

Checklist

• I kept the change scoped to the relevant area.
• I am opening this from a separate branch, not pushing directly to main.
• I updated docs, examples, or setup notes when behavior changed.
• I added or updated tests when the change affects behavior. (N/A — docs only)
• I did not commit secrets, .env files, dependency folders, or generated output.
• Active relative links in Markdown files resolve.

Notes for Reviewers

Worth deciding explicitly: should this also be a GitHub Security Advisory?
Right now the only public statement of the 1.2.0 exposure is this changelog
entry plus the v1.2.1 release notes. Without a GHSA, Dependabot will not notify
anyone still running 1.2.0. The repository has one prior advisory,
GHSA-c795-2g9c-j48m, for the sibling sender_id traversal — same class, same
containment approach — so there is precedent for filing one here.

If a GHSA is filed, the affected range needs both segments (< 1.1.4 and
= 1.2.0); collapsing it to < 1.2.1 would incorrectly flag 1.2.0 as the
only safe pre-fix release... and would wrongly mark 1.1.4 users as vulnerable.

By submitting this pull request, I agree that my contribution is licensed under
the Apache License 2.0.