feat(temples): make recovery pricing need-aware

[zoeyrose]

Status

Draft and blocked pending the work-aware service contract in
atrinik/classic#264. The current
head is a validated prototype, but it still contains the superseded hard-coded
provider registry, exact-price quote, pre/postcondition inference, and no
provider-credit support.

Revised target

• author a unique canonical npc_id and custom key/value
  temple_service_rank on every provider NPC;
• remove map/name/provider dispatch from Python and validate the authored
  identity/rank inventory;
• query bounded named spell-work components and convert them into a
  deterministic minimum/maximum monetary quote;
• bind state evidence, provider identity/rank, price range, existing debt, and
  maximum possible new debt into confirmation;
• re-query on confirmation and re-quote any drift without casting;
• calculate the final charge from the actual beneficial work reported by the
  cast, with zero cash and zero debt for zero work; and
• when explicitly confirmed work exceeds available funds, collect the
  available amount and record the bounded interest-free remainder against that
  provider's npc_id.

An indebted provider requires full repayment before another paid service.
Other providers are independent, and eligible newcomer-essential care remains
free and never creates debt. The actual cash plus debt can never exceed the
confirmed maximum.

Dependencies

• atrinik/classic#265 — native
  side-effect-free work ranges;
• atrinik/classic#266 — Python
  work-query API;
• atrinik/classic#248 — actual
  work outcomes; and
• atrinik/classic#267 — bounded
  provider-specific credit and repayment.

The same APIs will support future per-item Smith identification pricing under
#217, which does not block this
temple PR.

Required validation after revision

• python3 tools/validate.py and git diff --check
• complete provider npc_id/rank and save-enabled-map audit
• exact/ranged quote, actual-work, zero-work, partial-funding, debt, and
  repayment unit/runtime coverage
• derived Classic server build and complete CTest suite
• isolated low/high patient and provider scenarios with reconnect/cleanup
• fresh independent whole-diff reviews with zero actionable findings
• latest-head required PR checks

Superseded current-head evidence

Head 1dce4f43ebe4d5d8d5582d3f06f395da70254455 previously passed full content
validation, the derived Classic test suite, an isolated server scenario, two
whole-diff reviews, and all required checks. Those results describe only the
superseded prototype and are not acceptance evidence for the revised design.

Closes #106

[zoeyrose]

Delivery review complete at 1dce4f43ebe4d5d8d5582d3f06f395da70254455.

• Full content validation passed (174 tests plus provider, map, schema, catalog, syntax, resource, and licensing stages).
• Derived Classic server validation passed all 48 CTest targets; the 441-test Python plugin run includes five live temple pricing/treatment cases.
• Isolated issue-106-temple-pricing / basic-player scenario reached server-ready state and was shut down.
• Two fresh post-fix whole-diff reviews completed with zero actionable findings.
• All four latest-head GitHub checks passed.

The nonblocking native Cast() outcome follow-up remains tracked in atrinik/classic#248.

[zoeyrose]

This PR is paused for the review-directed dependency work tracked by
atrinik/classic#264 and has been
returned to draft.

Before it becomes ready again, the implementation will:

• replace the hard-coded provider registry with an authored
  temple_service_rank key/value attribute on each provider NPC;
• use the native side-effect-free query to reject services that cannot change
  the target under the provider's authored rank;
• settle from the actual Cast()/CastIdentify() result so probabilistic and
  deterministic no-ops are never charged; and
• rerun content validation, derived Classic runtime coverage, latest-head PR
  checks, and fresh whole-diff reviews.

The current successful checks describe the existing head only and are not the
acceptance evidence for the revised implementation.

[zoeyrose]

Scope amendment recorded: this draft will move from one exact pre-cast price to
a bounded work-based range and final actual-work settlement.

The eventual confirmation will show the range, available funds, existing debt,
and maximum possible new debt. A successful cast charges only for reported
beneficial work; any unaffordable remainder is recorded—only with explicit
consent—against the provider's stable npc_id. Zero work remains free.

The PR body now lists the revised Classic dependencies (#265, #266, #248,
#267), the authored metadata changes, and the complete validation/review cycle
required before it can return to review. Identification adoption is tracked in
#217 without blocking this temple PR.