Make trio + telco templates compatible with PyRel 1.11

[cafzal]

Bumps factory_production, supplier_reliability, cicd_runner_allocation, and telco_network_recovery to relationalai==1.11.0 and fixes the incompatibilities that surfaced.

Changes

• Strip mismatched short_name= — PyRel 1.11 rejects a short_name= that differs from its LHS attribute at model-build ([Invalid short name]). Removed the 5 mismatched args in telco_network_recovery.py (performance_for_tower, cdr_caller, cdr_callee, cdr_routed_through, upgrade_for_tower); PyRel now derives each name from the attribute. Each value was used only in its own declaration. Every other short_name= in v1/ already matches its attribute and is left untouched.
• Fix solve() signature — Problem.solve() now requires solver as a positional arg. telco's problem.solve() (which relied on the removed engine-default) → problem.solve("gurobi"), preserving the existing HiGHS fallback. The trio already used solve("highs", ...).
• Pins — trio (==1.9.0) and telco (==1.8) → ==1.11.0; README SDK requirement lines aligned.

Validation

All four run end-to-end on relationalai==1.11.0, no [Invalid short name] errors:

• telco — PIPELINE COMPLETE, plan $4,998,891 (24 towers, binding=budget)
• factory_production — OPTIMAL, profit 200000 + sensitivity
• supplier_reliability — OPTIMAL, obj 4850 + scenario analysis
• cicd_runner_allocation — OPTIMAL baselines + intentional CONFLICT_FOUND/IIS demo

[github-actions]

The docs preview for this pull request has been deployed to Vercel!

✅ Preview:	https://relationalai-docs-nuet26myl-relationalai.vercel.app/build/templates
🔍 Inspect:	https://vercel.com/relationalai/relationalai-docs/81in4RUc11uWidFZspEYXg3iyuCw

[chriscoey]

Approving — clean, well-scoped 1.11 bump. I dug into the two real changes and ran the templates to confirm.

Verified

• The short_name= removals are safe: in 1.11.0 relations are keyed by object identity, not by name string, so NetworkPerformance.for_tower and TowerUpgradeOption.for_tower both deriving the "for_tower" display name don't collide. The five removed values aren't referenced anywhere else in v1/.
• solve() in 1.11.0 is def solve(self, solver: str, *, …) — solver is positional-or-keyword, so the except fallback solve(solver="highs") is valid.
• Ran the trio end-to-end on relationalai==1.11.0: factory OPTIMAL, profit 200000 + sensitivity; supplier OPTIMAL, obj 4850 + scenarios (-C=6750, -B=5150); cicd OPTIMAL baselines (0.5x/1.0x/1.5x = $10.18/$9.62/$9.53) + the intentional CONFLICT_FOUND/IIS. All match the PR description. I didn't stand up telco's full GNN pipeline locally (needs change-tracking + EXP_DATABASE), so I'm relying on your telco run for Stage 4 — its model build passes here, so the short_name change is fine.

Non-blocking nits (fine to merge as-is):

1. telco_network_recovery/README.md:57 — the prerequisite bullet still says HiGHS fallback happens "if gurobi is not configured." With the new solve("gurobi"), fallback only fires when the engine raises a matching error, not when the engine is simply configured for HiGHS. The troubleshooting block at line 369 was updated to match; this earlier bullet reads stale. Maybe: "if gurobi is unavailable or unlicensed, the script catches the solver error and falls back to HiGHS automatically."
2. The fallback filter (telco_network_recovery.py:892) keys on the substring "unavailable" but not "not available", and it now runs on every gurobi-unconfigured solve rather than only genuine license failures — and it's the one branch neither run exercised (yours solved on gurobi, my trio runs don't touch telco). Worth a glance that your engine's gurobi-missing message actually matches the keyword set, else the else: raise hard-fails instead of falling back.
3. Minor: the comment at telco_network_recovery.py:882 ("engine-configured for licensed deployments") reads slightly stale now that the call passes "gurobi" literally rather than deferring to engine config.
4. Optional: the inline solve(sensitivity=True) / solve(conflict=True) shorthand in the trio READMEs isn't literally runnable under 1.11 (missing the now-required solver) — reads fine as illustrative shorthand, your call whether to tighten.

Nice cleanup.