fix: report version (#45) + re-run docs & flexible --report-dir (#48)

[NikolaosSokos]

Summary

Two focused fixes from the #45–#50 batch:

#45 — Tool version in reports

• Add eida_consistency.__version__ (sourced from installed package metadata, so it tracks pyproject.toml automatically).
• Surface it as summary.version in the JSON report and a Tool version line in the markdown Run Summary.

#48 — Re-run docs + flexible --report-dir

• consistency and explore now accept --report-dir at the subcommand level too (not only as a global option), with the local value taking precedence. The natural eida-consistency consistency … --report-dir DIR order no longer fails with a usage error — the root cause of the issue.
• Documentation (README.md, docs/cli.md):
  • Fix wrong --report-dir placement and the explore examples (--latest → no-arg form).
  • Correct the report filename pattern to <node>_<YYYYMMDD_HHMMSS>_<seed>.
  • Add a single canonical Re-run & Re-verify section (explore / check / fresh run) and remove the duplicated explanations.
  • Describe --seed realistically (does not reproduce an old run; replay the exact window instead).
  • Mention the published docs site in the README (badge + intro line).

Note: #46 was investigated and found to be a real node-side inconsistency (EPOSFR availability over-claims vs dataselect 204), not a tool false positive — no code change. Seed removal is tracked separately (#52).

Test plan

• uv run pytest — 164 passed
• New tests: --report-dir accepted both before and after the subcommand (consistency + explore)
• New tests: summary.version in JSON + Tool version line in markdown
• mkdocs build --strict passes

Closes #45
Closes #48