[Oaklight]

Why

CI runs bare pytest, so the SSE suite is covered there — but the documented
local workflow is make test, and sse was never added to the Makefile. A
contributor following the project docs silently skipped 67 tests. The two
modules SSE composes with (httpclient for transport, httpserver for serving)
also never mentioned it, so the pairing was only discoverable by reading the
module index.

What

• Makefile: test-sse runs test_sse_correctness.py and
  test_sse_inject.py with -rs; benchmark-sse runs the httpx-sse
  comparison. Both are in the .PHONY list, the test/benchmark
  aggregates, and make help.
• httpclient docs (EN + ZH): the streaming section now shows
  EventSource(r.iter_lines()) for text/event-stream instead of leaving
  line-splitting to the reader.
• httpserver docs (EN + ZH): the SSE example now uses encode_event() /
  encode_comment() and explains why hand-writing data: ...\n\n breaks on
  payloads containing a blank line.
• Changelog (EN + ZH): infrastructure entry for the missing targets.
• Module index config: the sse description now mentions the event encoder,
  matching the regenerated modules/index.md on both docs branches.

Verification

• make test-sse → 67 passed.
• make benchmark-sse → 6 passed.
• make help lists both targets.

The docs live on the docs_en / docs_zh worktree branches per project
convention and are pushed separately.

Fixes #242