feat(agent-reach): add LoongSuite instrumentation for Agent-Reach

Implements OpenTelemetry instrumentation for the Agent-Reach framework
(https://github.com/Panniantong/Agent-Reach) per the execute.md plan in
llm-dev/agent-reach/investigate/.

Coverage:
- ENTRY span on agent_reach.cli.main
- TOOL span on agent_reach.doctor.check_all
- TOOL span on every Channel.check override (dynamic discovery via
  register_post_import_hook on agent_reach.channels)
- TOOL span on agent_reach.probe.probe_command
- TOOL span on agent_reach.transcribe.transcribe and transcribe_chunk
- Conditional TOOL span on agent_reach.integrations.mcp_server.create_server
  (skipped when loongsuite-instrumentation-mcp is already active to
  avoid duplicate spans)

Spans are produced via opentelemetry-util-genai's ExtendedTelemetryHandler
(entry/execute_tool context managers). Sensitive content
(gen_ai.tool.call.arguments / .result) is gated behind
OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT plus the experimental
semconv opt-in, matching the util-genai contract.

Custom metrics (agent_reach_channel_status, agent_reach_probe_total,
agent_reach_probe_duration_seconds) record channel health and probe
outcomes; generic request/error/duration metrics are derived by the
backend from span data.

All 7 unit tests pass; ruff lint clean.

Co-authored-by: multica-agent <github@multica.ai>

Author: 123liuziming

instrumentation-loongsuite/loongsuite-instrumentation-agent-reach/README.md

@@ -0,0 +1,44 @@
+# LoongSuite Agent-Reach Instrumentation
+
+OpenTelemetry instrumentation for [Agent-Reach](https://github.com/Panniantong/Agent-Reach)
+(`pip install agent-reach`).
+
+## Coverage
+
+| Span | Operation | Patch point |
+|------|-----------|-------------|
+| ENTRY | `enter` | `agent_reach.cli.main` |
+| TOOL | `execute_tool` | `agent_reach.doctor.check_all` |
+| TOOL | `execute_tool` | every `Channel.check` override (dynamic discovery via `register_post_import_hook`) |
+| TOOL | `execute_tool` | `agent_reach.probe.probe_command` |
+| TOOL | `execute_tool` | `agent_reach.transcribe.transcribe` |
+| TOOL | `execute_tool` | `agent_reach.transcribe.transcribe_chunk` |
+| TOOL | `execute_tool` | `agent_reach.integrations.mcp_server.create_server` (only when `loongsuite-instrumentation-mcp` is NOT active) |
+
+Span hierarchy:
+
+```
+ENTRY (enter_ai_application_system)
+  ├─ TOOL (execute_tool doctor)
+  │    ├─ TOOL (execute_tool channel-github)
+  │    │    └─ TOOL (execute_tool probe-gh)
+  │    ├─ TOOL (execute_tool channel-twitter)
+  │    │    └─ TOOL (execute_tool probe-twitter)
+  │    └─ … (remaining channels)
+  ├─ TOOL (execute_tool agent-reach-transcribe)
+  │    └─ TOOL (execute_tool whisper-transcribe) × N
+  └─ TOOL (execute_tool mcp-get_status) [conditional]
+```
+
+Sensitive content (`gen_ai.tool.call.arguments` / `gen_ai.tool.call.result`)
+is gated behind `OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT`. When
+the flag is unset / `NO_CONTENT`, only non-PII diagnostic attributes
+(`agent_reach.channel.status`, `agent_reach.probe.status`, etc.) are emitted.
+
+## Usage
+
+```python
+from opentelemetry.instrumentation.agent_reach import AgentReachInstrumentor
+
+AgentReachInstrumentor().instrument()
+```

instrumentation-loongsuite/loongsuite-instrumentation-agent-reach/pyproject.toml

@@ -0,0 +1,66 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "loongsuite-instrumentation-agent-reach"
+dynamic = ["version"]
+description = "LoongSuite Agent-Reach Instrumentation"
+readme = "README.md"
+license = "Apache-2.0"
+requires-python = ">=3.10"
+authors = [
+  { name = "Ziming Liu", email = "liuziming.lzm@alibaba-inc.com" },
+  { name = "OpenTelemetry Authors", email = "cncf-opentelemetry-contributors@lists.cncf.io" },
+]
+classifiers = [
+  "Development Status :: 4 - Beta",
+  "Intended Audience :: Developers",
+  "License :: OSI Approved :: Apache Software License",
+  "Programming Language :: Python",
+  "Programming Language :: Python :: 3.10",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Programming Language :: Python :: 3.13",
+]
+dependencies = [
+  "opentelemetry-api ~= 1.37",
+  "opentelemetry-instrumentation >= 0.58b0",
+  "opentelemetry-semantic-conventions >= 0.58b0",
+  "opentelemetry-util-genai",
+  "wrapt >= 1.17.3, < 2.0.0",
+]
+
+[project.optional-dependencies]
+instruments = [
+  "agent-reach >= 1.5.0",
+]
+
+test = [
+  "pytest ~= 8.0",
+  "pytest-asyncio ~= 0.23.0",
+  "pytest-cov ~= 4.1.0",
+  "agent-reach >= 1.5.0",
+]
+
+[project.entry-points.opentelemetry_instrumentor]
+agent_reach = "opentelemetry.instrumentation.agent_reach:AgentReachInstrumentor"
+
+[project.urls]
+Homepage = "https://github.com/alibaba/loongsuite-python-agent/tree/main/instrumentation-loongsuite/loongsuite-instrumentation-agent-reach"
+Repository = "https://github.com/alibaba/loongsuite-python-agent"
+
+[tool.hatch.version]
+path = "src/opentelemetry/instrumentation/agent_reach/version.py"
+
+[tool.hatch.build.targets.sdist]
+include = [
+  "/src",
+  "/tests",
+]
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/opentelemetry"]
+
+[tool.pytest.ini_options]
+asyncio_mode = "auto"