feat(#149): add resource ceiling assertions to bench harnesses

[Keengfk]

feat(#149): add resource ceiling assertions to bench harnesses

Summary

Adds hard CPU and memory budget ceilings to every public entrypoint in
intent_settlement and solver_registry. A test fails immediately in CI whenever
a code change causes an entrypoint to exceed its ceiling, making silent
resource regressions impossible.

What changed

intent_settlement/src/bench.rs (rewritten)

• Ceiling constants (CPU_CEIL_* / MEM_CEIL_*) for all 12 single-call
  entrypoints: submit_intent, accept_intent, fill_intent (full + partial),
  cancel_intent, expire_intent, slash_solver, request_extension, register_solver
  (first + top-up), withdraw_bond, deregister_solver
• Worst-case batch ceilings for batch_submit_intent, batch_accept_intent,
  batch_fill_intent, batch_cancel_intent at MAX_BATCH_SIZE = 20
• One bench_ceiling_* test per entrypoint using an assert_within_ceiling helper
• Existing resource_cost_report and resource_cost_is_reproducible tests
  retained unchanged

solver_registry/src/bench.rs (new file)

• Full bench harness matching the intent_settlement pattern
• Ceiling assertions for all 22 public entrypoints: initialize, set_writer,
  set_tier_threshold, register_solver, stake, unstake, deregister_solver,
  record_fill, record_failure, slash, get_tier, tier_for, get_reputation_score,
  get_solver, get_solver_count, get_tier_table, get_fill_window_bonus_pct,
  get_slash_bps, get_fee_rebate_bps, get_admin, get_bond_token,
  compute_reputation_score
• resource_cost_report and resource_cost_is_reproducible smoke tests included

solver_registry/src/lib.rs

• Added #[cfg(test)] mod bench; module declaration

Methodology

Ceilings are set at baseline + 10% (rounded up to the nearest 1 000). Baselines
for intent_settlement come from the existing
docs/149-resource-cost-per-entrypoint.md snapshot (soroban-sdk 21.7.7).
Baselines for solver_registry are first-capture values at the same SDK version.

To update ceilings after a deliberate cost change:

1. Run cargo test --features testutils bench::resource_cost_report --
   --nocapture to get new measured values
2. Set CPU_CEIL_* / MEM_CEIL_* to new_value * 110 / 100
3. Update docs/149-*.md with the new baseline

Out of scope

Cost optimisation is tracked separately. This PR only adds the guard rails.
closes #423

[drips-wave]

@Keengfk Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits