fix(viewer): skip unreadable/non-UTF-8 files instead of aborting the whole bundle

[mftnakrsu]

What

enrichment_agent visualize aborts the entire bundle if a single concept file can't be read, decoded, or parsed.

Why

In viewer/generator.py, _walk_concepts reads every concept with read_text(encoding="utf-8") and parses it, but only caught OKFDocumentError. Three realistic failure modes still escaped and killed the whole visualize command:

• invalid UTF-8 bytes → UnicodeDecodeError
• an unreadable file → OSError
• a malformed frontmatter value — e.g. an out-of-range timestamp: 2026-13-45 — makes PyYAML's implicit resolver raise a bare ValueError that OKFDocument.parse does not wrap

One bad file takes down the whole graph. The sibling bundle/index.py::_load_doc already tolerates all of these (except Exception: return None).

Change

• Skip the offending file by catching (ValueError, OSError). ValueError subsumes OKFDocumentError and UnicodeDecodeError (both subclasses) as well as PyYAML's bare ValueError; OSError covers read errors. The now-unused OKFDocumentError import is dropped.
• Add two regression tests: test_unreadable_concept_file_is_skipped (invalid UTF-8) and test_malformed_timestamp_concept_file_is_skipped (bad timestamp → bare ValueError), each asserting the remaining concepts still render.

Tests: pytest okf/tests/test_viewer.py → 8 passed.

🤖 Generated with Claude Code

[mftnakrsu]

@amirhormati a small viewer robustness fix — skip unreadable/non-UTF-8 files instead of aborting the whole visualize run. Rebased onto the latest main; checks are green and it's mergeable. Would appreciate a review when you get a chance. Thanks!