Give the shop a daily money budget that recharges (closes #20)

[dmccoystephenson]

Summary

Implements #20 ("make the shop's money recharge over time").

Design decision (made per request): the budget gates selling — otherwise shop money is meaningless flavor — but it's tuned to be forgiving.

• src/location/shop.py: the shop holds a daily budget (SHOP_DAILY_BUDGET = 750) that refills lazily at the start of each new day (_refillIfNewDay, keyed off timeService.day — no TimeService coupling).
• sellFish buys your fish most-valuable-first until the budget runs out; unaffordable fish stay in your inventory, with a clear "shop is out of money for today — come back tomorrow" message. A normal catch sells cleanly; only large hauls hit the cap.
• No persistence/schema change (the shop isn't saved) — budget starts full and refills daily. Removed the now-unused random import.
• README: brief "Selling Fish" note on the daily budget.

Balance (playtest simulation on the real game code)

At $750/day: optimal reinvest slows only mildly (~20 → ~24 days to $10k), the steady-income business strategy stays competitive (~25), and plain grind is ~113. The budget adds a real "sell regularly / bank your wealth" constraint without crippling skilled play (checked $500/$750/$1000 and picked $750).

Test plan

• python3 -m compileall -q src clean
• SDL_VIDEODRIVER=dummy SDL_AUDIODRIVER=dummy python3 -m pytest — 224 passed
• New tests: budget gates a large haul (leftovers + "out of money" message), budget refills next day, and the no-fish message. Existing sell tests (normal catch, legacy aggregate, sell-by-species) still pass under the budget.

Known limitation

The shop's budget isn't persisted, so it resets to full on reload — but since a normal Sleep already refills it (plus restoring energy and paying interest/crew), reload-to-refill is strictly worse than just sleeping and isn't a useful exploit. Persisting it would require reintroducing shop save state; noted for a possible follow-up.

Closes #20

🤖 Generated with Claude Code

[dmccoystephenson]

Self-review:

• Scope: PASS — shop.py + its tests + a README note; all serve #20.
• Tests-fix/new: PASS — budget-gating/refill/no-fish tests; the gating test fails under the old sell-everything code; existing sell tests (normal/legacy/by-species) still pass.
• Schema/save-safety: PASS — shop isn't persisted, so no schema/reader-writer change; not a protected path.
• Design call: gating chosen deliberately and tuned via the playtest sim ($750 → reinvest ~24, business ~25); documented + known reload limitation noted.
• Clean-up: removed the now-unused random import. Money-format: messages use %.2f. CI: PASS.