S24.04: Migrate Linux BDD off Example/SingleTask; remove the SingleTask binary

[DavidCozens]

Parent epic: #254 (E24: Code Hygiene)

Overview

Example/SingleTask/ (Linux, NullBuffer, direct UDP/TCP send) and
Example/Threaded/ (Linux, PosixMessageQueueBuffer + service thread,
SwitchingSender over UDP/TCP/TLS/mTLS, BlockStore option) are two
binaries doing overlapping jobs. The Threaded binary is a strict
superset since S15 introduced SwitchingSender. The BDD harness uses
both — most scenarios run against the SingleTask binary via
run_example; the @buffered-tagged scenarios route to the Threaded
binary via run_buffered_example. Windows already collapsed to one
binary in S13.20; FreeRTOS only has one (the buffered SingleTask after
S08.04 slice 2). Linux is the only platform left carrying two.

Why

• Two binaries per platform was historical (S03 era); SwitchingSender
  (S15) made the second one redundant for BDD coverage.
• The Linux SingleTask binary is BDD test infrastructure dressed as a
  pedagogical example. Pre-launch (v0.0.0) there are no external
  readers to consider; the duplication is pure carrying cost.
• Future SL1 / SL2 / SL3 pedagogical examples (separate epic, not
  scheduled here) want a clean Example/ directory to live in.

In scope

• Point BDD_TARGET=linux's context.example_binary default at the
  Threaded binary (build/debug/Example/SolidSyslogThreadedExample).
• Audit every run_example caller on Linux (currently ~25 scenarios
  via the the example program sends ... step family) to confirm none
  rely on NullBuffer's synchronous-send timing semantics. The
  Threaded binary's service-thread drain takes a few hundred ms in the
  worst case; any scenario asserting "oracle received within N ms of
  the send command returning" without already waiting for receipt is a
  surface we need to fix.
• Collapse run_example / run_buffered_example if the dispatch no
  longer needs to branch — they now both route to the Threaded binary
  on Linux.
• Delete Example/SingleTask/ (directory + add_subdirectory in the
  parent Example/CMakeLists.txt).
• Delete Tests/Example/SolidSyslogExampleTest.cpp and its
  add_subdirectory registration.
• Update Example/CMakeLists.txt and the top-level CMakeLists.

Acceptance criteria

• All Linux BDD scenarios that pass today continue to pass after
  the cut. Specifically bdd-linux-syslog-ng is green on CI.
• Example/SingleTask/ no longer exists.
• Tests/Example/SolidSyslogExampleTest.cpp no longer exists.
• analyze-format, analyze-tidy, analyze-cppcheck,
  coverage-linux-gcc all green (coverage stays ≥ 90%).
• No references to SolidSyslogExample (the SingleTask binary
  name) remain in the repo apart from blog-history mentions.

Out of scope

• Renaming or moving the Threaded binary out of Example/ — that's
  S24.05.
• Windows / FreeRTOS BDD wiring — they already use a single binary
  each.
• New SL1 / SL2 / SL3 pedagogical examples — separate future epic.
• Documentation rewrite — David is rewriting the docs separately;
  light touch only where needed to keep references valid.

Risks

• Some BDD steps may have implicit synchronous-send assumptions. The
  audit upfront is the gate.
• The THREADED_BINARY constant in Bdd/features/steps/syslog_steps.py
  becomes the only binary; the routing logic in run_buffered_example
  / _run_with_prompt_protocol simplifies accordingly.

References

• E24 E24: Code Hygiene #254 charter — "Other code-quality work that doesn't belong
  under a capability-focused epic".
• S13.20 — Windows already-collapsed precedent.
• S08.04 S08.04: FreeRTOS mutex — CircularBuffer for concurrent task emission #269 — FreeRTOS now ships a single buffered binary.
• Conversation 2026-05-11 in PR feat: S08.04 SolidSyslogFreeRtosMutex + buffered FreeRTOS example #330 close-out turn (post-S08.04
  merge) — wider plan David sketched, captured in agent memory
  project_example_consolidation_plan.