fix(compose): surface per-service scan failures in user-facing summary

[myukitty]

Summary

Closes #131

When scanning Docker Compose setups containing multiple services, if one or more services encounter scan execution failures (e.g. missing Dockerfile build context, daemon errors, or unresolvable targets), the failures were logged but not clearly surfaced in the user-facing quick take summary or structured JSON output.

This change:

1. Tracks and deduplicates per-service failures: \ComposeOrchestrator.run_full_scan\ collects \ailed_services\ (service name and failure reasons) and \ otal_services\ count.
2. Surfaces scan failures in Quick Take and JSON:
   • In _quick_take_lines: Appends \"{N} of {total} services had scan failures: {service_names}"\ when failures occur.
   • In _print_json_results: Includes \ailed_services\ and \ otal_services\ in \scan_info\ when failures are present.
3. Robust diagnostics classification: Correctly distinguishes Hadolint lint diagnostics from tool execution errors so lint warnings are presented as lint counts rather than tool failures.
4. Early validation & path safety: Validates services mappings before scanner invocation and resolves relative build contexts relative to the Compose file directory.

Testing

• Added comprehensive unit tests in \ ests/test_compose_scanner.py\ covering failure tracking, deduping, Hadolint diagnostics classification, parse/malformed-services handling, and summary/JSON formatting.
• Full test suite passed (228/228 tests passing).