Keep segments within the MSS announced by the peer

[FrenchToblerone54]

Problem

Outgoing segments are cut to the stack's MTU and the peer's window only. The MSS option in the peer's SYN is ignored, so a peer behind a smaller link gets segments larger than it said it can take, which can be dropped or fragmented on the way (RFC 9293 section 3.7.1 says a sender must not send a segment larger than the MSS the peer announced).

Reproduce

Open a connection with a SYN carrying MaximumSegmentSize(600) and write 1460 bytes. The stack sends 1460 bytes in one segment. The new test does this.

Fix

• Tcb keeps peer_mss, unbounded by default so connections without the option behave as before.
• The option is read from the SYN only, like the window scale option.
• calculate_payload_max_len bounds the payload by the MSS minus the TCP options of the segment (the MSS counts options but not the 20-byte base header).
• A value below 88 is read as 88 (Linux's TCP_MIN_MSS), so a peer cannot force a few bytes per packet. RFC 9293 section 3.7.1 only fixes the default of 536 for a SYN without the option, it sets no floor for an announced value.

Test

segments_stay_within_the_peers_mss in src/stream/tcp.rs covers an announced MSS of 600 (segments of 600), of 100 (segments of 100) and of 40 (segments of 88). It fails without the change (left: 1460, right: 600) and passes with it. cargo test and cargo clippy --all-targets are clean.