Ensure TLS roles are respected during session resumption

[randombit]

An application which acted as both a TLS server and a TLS client, and which shared a session manager between those two roles, could end up accepting a resumed session that resumed in the wrong role.

[Copilot]

Pull request overview

This PR hardens Botan’s TLS session resumption logic against cross-role session confusion when a single Session_Manager instance is shared between a TLS client and TLS server in the same application.

Changes:

• Enforce connection-side (“role”) checks when resuming sessions on the server side (TLS 1.2 check_for_resume() and TLS 1.3 PSK selection paths).
• Filter client-side session discovery so that client handshakes won’t be offered server-established sessions.
• Add/adjust tests to ensure sessions cannot cross the client/server boundary (including PSK resumption cases) and adapt RFC 8448 server tests accordingly.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/tests/test_tls_session_manager.cpp	Adds multi-Session_Manager regression tests ensuring role separation and updates existing session-manager tests to match new semantics.
src/tests/test_tls_rfc8448.cpp	Adjusts RFC 8448 server-side resumption tests to use a server-role Session variant.
src/lib/tls/tls13/tls_extensions_psk.cpp	Adds a role check so the server won’t resume using client-role sessions via the PSK resumption path.
src/lib/tls/tls12/tls_server_impl_12.cpp	Adds a server-role guard when checking for TLS 1.2 resumption.
src/lib/tls/tls_session_manager.h	Documents the new role-based behavior of retrieve() and find().
src/lib/tls/tls_session_manager.cpp	Implements role filtering in retrieve() and find_and_filter() to prevent cross-role session use.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[reneme]

In fact, I would expand on the proposed changes (perhaps in a follow-up), like so:

1. Extend the documentation of the virtual storage interface methods: It already states that retrieve_one is only called on the server side and likewise find_some only on the client side. Those docstrings should also mention that the methods must filter their outputs accordingly (making the filters introduced in this PR a belt-and-suspenders idiom ensuring that custom session managers don't get funky).
2. Adapt our TLS::Session_Manager implementations: we provide a bunch of them, and they should filter their outputs accordingly.

As a result, applications that hold both client and server state in the same place should have a much higher chance of actually benefitting from resumption as expected.

[randombit]

Review comments addressed and I extended the SQL and memory session managers to filter their results