fix(server): decode streamed SSE text from the cumulative id list

[ahrazzle]

What changed

Streamed SSE responses now decode the full id list cumulatively and emit only the newly completed suffix, holding back partial multi-byte characters until they resolve.

Root cause

The streaming path decoded each token id in isolation (python/src/edge0/server/app.py:82). A byte-level BPE token carrying part of a multi-byte character streamed as U+FFFD while the non-streaming path returned correct text.

Changes

• Added _incremental_suffix helper to extract the new suffix from a cumulative decode
• Modified _chat_stream to maintain a running list of seen ids
• Flush held-back partials at end of generation so streamed equals non-streaming text

Rebase needed

The multi-platform layout refactor (cad29a1) moved the touched files to python/src/edge0/server/app.py and python/tests/test_server.py. This branch predates it and no longer applies (mergeable: conflicting). Rebase onto main first. The diff carries over unchanged: a local rebase onto feafe31 applied with rename detection and zero content conflicts.

Validation

Live edge0-8b server run (stdlib transport), prompt asking the model to echo 🌊🏖️🦀🍣:

path	text
base, streaming	9 U+FFFD, no emoji
head, streaming	🌊🏖️🦀🍣 (0 U+FFFD)

The non-streaming path decodes the whole id list and was already correct.

Test: python/tests/test_server.py::test_chat_stream_incremental_decode_of_split_characters

• Red on base (base tree + head test file)
• Green on head (66 passed, 1 skipped)
• Re-verified 2026-10-03 on the rebased tree: test passes (34 ids, 14 content deltas, streamed text == non-streamed text, no U+FFFD). The same test fails on unpatched base

Limits

Common CJK text in the shipped vocabulary is not affected. Corruption needs characters that split across byte tokens (emoji, rare CJK, some Korean).

O(n²) cumulative re-decode per request. This is the minimal fix. Throughput was not measured.

Automated posting by agentic team with human oversight.