test: add SDK-focused integration tests and expand CLI smoke

[dapi]

Summary

Current CI validates unit tests and a narrow CLI smoke path (send --parse-mode).
For this project, we need stronger confidence in SDK-level behavior and real command wiring.

Why

• Most tests are mocked unit tests around telegram-client methods.
• No integration suite that validates MCP SDK tool wiring end-to-end in a controlled environment.
• CLI smoke currently checks only parse-mode validation and misses core command surfaces.

Scope

1. Add an SDK-focused integration test suite.
2. Expand CLI smoke checks in CI with deterministic, non-network scenarios.

Proposed implementation

1) Integration tests (new tests/integration/)

• Start MCP server in test mode with temporary TGCLI_STORE.
• Verify MCP tool registration and call path for:
  • listChannels
  • searchChannels
  • getChannelMessages
  • scheduleMessageSync
  • listMessageSyncJobs
• Use controlled mocks/fixtures for MTProto boundaries, but keep server + tool execution real.
• Validate response schemas and error contracts.

2) CLI smoke expansion (CI)

• Keep existing parse-mode checks.
• Add deterministic smoke checks for:
  • tgcli --help
  • tgcli <group> --help for key command groups (sync, messages, channels, send, server, config)
  • invalid argument handling and exit codes
  • temp-store config read/write (TGCLI_STORE=$(mktemp -d))
• Ensure no external Telegram network dependency in smoke job.

3) CI structure

• Split test scripts:
  • npm run test:unit
  • npm run test:integration
  • npm run test:smoke
• Update CI workflow to run all three and fail fast on contract regressions.

Acceptance criteria

• Integration tests run in CI and validate MCP SDK tool execution path.
• CLI smoke covers all major command groups + exit code expectations.
• No integration/smoke step requires real Telegram account or network login.
• CI stays under an acceptable runtime budget (target: < 5 minutes total).

Notes

Primary quality gate should be SDK behavior correctness, not only method-level units.