chore: update tend workflows (0.1.15 → 0.1.16)

[prql-bot]

Nightly regeneration of tend's workflow files with uvx tend@latest init.

tend version: 0.1.15 → 0.1.16

Notable changes

• The tend-mention pre-check was consolidated (simplify: coarser mention gates, tested poll scripts, anchor-based run window max-sixty/tend#965, simplify: one tested review-anchor resolver, skill depth into references max-sixty/tend#971) — this is the only file with a body change; the other seven are action-ref and header-stamp bumps. The bot's login is now injected as a BOT_NAME env var rather than baked into each comparison, and the three separate bot-self-review skips collapse into one gate: the bot's own review summons a session only when it hands work from the reviewer role to the author role (a body, or a fresh non-reply inline comment, on a PR the bot authored).
• A contentless approval — no body, no inline comments — is now terminal regardless of who submitted it, where before only the bot's own such approval was skipped (simplify: coarser mention gates, tested poll scripts, anchor-based run window max-sixty/tend#965).
• The bundled skills now gate self-initiated fixes on cost as well as evidence (skills: gate self-initiated fixes on cost, not just evidence max-sixty/tend#960), so nightly and triage runs weigh the size of a fix before opening a PR.
• tend-weekly no longer trusts a re-anchored approval on a rebased dependency PR (fix(weekly): don't trust a re-anchored approval on a rebased dependency PR max-sixty/tend#890), and tend-review-runs line-anchors its evidence-log selector (fix(review-runs): line-anchor the evidence-log selector max-sixty/tend#963). Both are skill changes that arrive via the action-ref bump.

Full comparison: max-sixty/tend@0.1.15...0.1.16