fix(analyzer): align file-size guard with character semantics (#284)

[rodboev]

Summary

The analyzer file-size guard already measures decoded Python characters, but the shared constant name and skip logs still describe that threshold as bytes. Multibyte UTF-8 content therefore follows the current compatibility policy while being reported with the wrong unit.

Closes #284

Root cause

src/skillspector/nodes/analyzers/static_runner.py owns the shared analyzer size guard, and every affected analyzer compares len(content) on decoded str content against that constant. The value and behavior were already character-based; only the constant name, skip logs, and one adjacent _agent_cli.py comment still claimed byte semantics.

Diff Notes

• Renamed MAX_FILE_BYTES to MAX_FILE_CHARS in src/skillspector/nodes/analyzers/static_runner.py and updated static patterns, YARA, behavioral AST, and behavioral taint consumers to import the renamed authority.
• Updated static-runner and YARA oversize skip logs to report character units instead of bytes.
• Rewrote the _agent_cli.py cap comment so the analyzer character limit and the CLI's encoded-byte limit are described as separate guards.
• Added focused boundary, over-boundary, multibyte-under-limit, later-component-continuation, and log-wording coverage in tests/nodes/analyzers/test_static_runner_filtering.py, tests/nodes/analyzers/test_static_yara.py, tests/nodes/analyzers/test_behavioral_ast.py, and tests/nodes/analyzers/test_behavioral_taint_tracking.py.

Scope

This keeps the existing one-million-character analyzer policy and per-file skip behavior. It does not switch analyzers to encoded-byte limits, change build_context size_bytes metadata, touch input acquisition or report schemas, or alter _agent_cli.py runtime behavior.

Verification

• uv sync --all-extras passed; Resolved 156 packages in 0.88ms and Checked 151 packages in 3ms.
• cmd.exe /d /s /c "set PYTHONPATH=src&& .\.venv\Scripts\python.exe -m pytest tests\nodes\analyzers\test_static_runner_filtering.py tests\nodes\analyzers\test_static_yara.py tests\nodes\analyzers\test_behavioral_ast.py tests\nodes\analyzers\test_behavioral_taint_tracking.py" passed; 206 passed in 44.64s.
• uv run ruff check src/ tests/ passed.
• uv run ruff format --check src/ tests/ passed.
• git diff --check passed.

[koriyoshi2041]

The character-based policy is applied consistently across the four analyzer paths, and the renamed constant/log wording now matches the actual len(str) guard. The boundary and multibyte cases also preserve the existing behavior explicitly.

I ran the four focused analyzer test modules on this head: 206 passed.