W2a: #49-P1 substrate — generateUntilStop in llm-core, demo/CLI out of kllama, shared ModelMetadata extraction (#37) - #296

…ore, extract demo/agent CLI from kllama, shared ModelMetadata extraction (#37)

Three moves that finish decoupling tool calling from the kllama runner
(#49 Phase 1) and make capability metadata shared and testable (#37):

- generateUntilStop + GenerateResult + tensor sampleFromLogits move to
  llm-core's sk.ainet.apps.llm, next to InferenceRuntime/PrefillStrategy.
  Any runner now gets EOS-aware generation without depending on the agent
  layer. llm-agent keeps typealias/delegating re-exports (same pattern as
  the earlier InferenceRuntime promotion), so no caller changes.
- ToolCallingDemo, AgentCli, ListFilesTool, CalculatorTool move from
  :llm-runtime:kllama jvmMain to :llm-agent jvmMain, packages preserved.
  kgemma and skainet-cli previously imported these *from kllama* — the
  exact cross-runner coupling #49 Phase 1 calls out; both already depend
  on llm-agent directly and need no source changes.
- New ModelMetadataExtraction in llm-agent: pure, format-agnostic
  best-effort extraction (GGUF field map -> family/architecture/
  chat_template/tokenizerHints). kllama's private peekGgufMetadata now
  delegates to it; unit tests cover full/missing/malformed fields and
  that extracted metadata drives resolver auto-detection.

apiDump refreshed for llm-core, llm-agent, kllama; apiCheck green.
Part of the #35 epic; W2a of three stacked PRs (W2a/W2b/W2c).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

…tests, Qwen real-GGUF validation, native-vs-generic docs

Final PR of the W2 stack (#35 epic):

- #41: AgentCli (chat/agent modes) now prints the same provider/mode/
  reason resolution diagnostics ToolCallingDemo already printed, so
  native-vs-generic selection is visible in every CLI mode.
- #42: ResolutionDiagnosticsTest pins resolveWithDiagnostics — explicit
  override reason, auto-detect reason, GENERIC fallback mode/reason, and
  extraction→diagnostics carry-through. (GGUF/HF extraction tests landed
  in the earlier PRs of this stack.)
- #43: QwenToolCallSmokeTest, env-gated on QWEN_MODEL_PATH. Two stages:
  (1) metadata extraction + auto-resolution against the real file —
  header-only, runs in seconds; verified PASSING against
  Qwen2.5-0.5B-Instruct-F16.gguf (family=qwen, arch=qwen2, hints
  [<tool_call>, <tool_response>, <|im_start|>] → provider qwen, NATIVE,
  no explicit family). (2) full agent-loop round-trip over the real
  checkpoint via the same DSL path the CLI uses — currently fails for a
  reason outside this architecture: the DSL Qwen runtime emits
  degenerate text on real checkpoints (tracked as #118); documented in
  the test so it becomes the round-trip pin once #118 lands.
- #44: README gains a "Tool calling: native vs. generic support"
  section with resolution-order docs and a family/provider/mode/format/
  detection compatibility matrix; CHANGELOG entry for the W2 stack.

Verification: llm-agent jvmTest 210 green; kllama jvmTest green
(gated tests skip cleanly without env vars); apiCheck green on
llm-core/llm-agent/kllama (no public API changes in this PR);
gemma-iree jvmTest green (functiongemma provider registration).

Note: :llm-apps:kllama-cli / :llm-apps:skainet-cli fail dependency
resolution on develop (empty engine version on app classpaths),
pre-existing and unrelated — demo validation therefore ran through the
env-gated smoke tests.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

W2c: AgentCli diagnostics (#41), detection tests (#42), Qwen real-GGUF validation (#43), native-vs-generic docs (#44)