Improve CLI validation, metadata, and project documentation

[VIVAAN-DHAWAN]

Summary

• centralize the package version and keep GitHub User-Agent metadata current
• reject non-positive pull-request numbers before making API calls
• add regression coverage for CLI argument validation
• improve README onboarding, badges, roadmap, development, and security guidance

Verification

• python3 -m compileall -q src tests action scripts passed
• git diff --check passed
• pytest -q could not run in the execution environment because pytest and runtime dependencies are not installed, and the configured pip executable is unavailable

This keeps the existing deterministic triage behavior intact while improving release hygiene, input validation, and contributor usability.

[github-actions]

Diffly verdict: QUARANTINE

#5 · 6 files · 77 lines changed · checks: PENDING

Risk flags: NEW_DEPENDENCY (medium), NO_TEST_COVERAGE (medium), CHECKS_UNKNOWN (medium)

Risk flags and reasoning

• NEW_DEPENDENCY (medium): Adds or changes a dependency manifest or lockfile. — pyproject.toml
• NO_TEST_COVERAGE (medium): Changed production files have no obvious neighboring or repository test coverage. — README.md; pyproject.toml; src/diffly_cli/__init__.py; src/diffly_cli/cli.py; src/diffly_cli/github.py
• CHECKS_UNKNOWN (medium): Required status checks are missing, pending, or unavailable. — pending
• QUARANTINE because dependency changes expand the supply-chain and runtime surface.
• QUARANTINE because at least one changed production file lacks obvious test coverage.
• QUARANTINE because the affected pull request does not have a confirmed passing check result.

Blast-radius summary

File	Change	Symbols	Direct callers	Tests
README.md	+44/-1	<top-level changes>	—	—
pyproject.toml	+1/-1	<top-level changes>	—	—
src/diffly_cli/__init__.py	+3/-1	<top-level changes>	—	—
src/diffly_cli/cli.py	+12/-1	<top-level changes>	—	—
src/diffly_cli/github.py	+2/-1	<top-level changes>	—	—
tests/test_regressions.py	+9/-1	<top-level changes>	—	—

Deterministic triage is authoritative; any optional LLM explanation cannot change the verdict.