Docs/claude code integration

[AnuGuin]

Closes #1358

Purpose / Description

Adds the missing integration documentation page for Claude Code
(docs/integrations/claude-code.md). Claude Code is a first-class IDE
in Observal with full hook support, session parsing, and e2e tests, but
had no setup or reference documentation page for users to find.

Fixes

Fixes #1358

Approach

Created docs/integrations/claude-code.md sourced directly from the
implementation files rather than inferred — every path, event name, and
behavior is verified against:

• observal_cli/ide/claude_code.py — adapter scan logic, hook detection threshold
• observal_cli/ide_specs/claude_code_hooks_spec.py — hook spec version, exact hook JSON
• observal_cli/sessions/claude_code.py — JSONL discovery, cursor logic, subagent push
• observal-server/services/ide/claude_code.py — agent config generation, frontmatter format
• observal-server/schemas/ide_registry.py — config paths, scope labels, event map

The page covers all sections required by the issue:

• Overview of the incremental session push architecture
• Supported features table (agents, skills, hooks, MCP, subagents, model selection)
• Setup steps (observal doctor --patch, observal pull)
• Config file paths for both project and user scope
• Hook spec (exact JSON written to settings.json, full event map)
• Session push behavior (byte-cursor, subagent JSONL layout)
• Caveats (PYTHONPATH on install, hook detection threshold, MCP config path split, agent marker timing guard)

How Has This Been Tested?

Documentation-only change. Verified no broken links and that all file
paths, event names, and version constants match the current codebase.

Ran the registry and adapter test suites to confirm no regressions:
pytest tests/test_cli_ide_adapters.py -q → 59 passed in 0.73s
pytest tests/test_ide_registry.py -q → 56 passed in 0.23s
pytest tests/test_constants_sync.py -q → 16 passed in 0.21s

Checklist

• Descriptive commit message
• Self-review completed
• Tests run for affected area
• UI changes: N/A (documentation only)

AI Assistance

Was generative AI tooling used to co-author this PR?

• Yes(Please Specify the tool): Claude Sonnet 4.6 for quick revision of the working of cli and server files.
• Was the generated code manually reviewed and tested? Passed linting and dedicated pytest tests