Rebind TimeService to the loaded player and stats after save load

[dmccoystephenson]

Summary

• FishE.__init__ builds TimeService(self.player, self.stats) from the new-game defaults, then loads each save file only if it exists and is non-empty. loadPlayer()/loadStats() rebind self.player/self.stats to brand-new objects, but only loadTimeService() rebuilds the TimeService around them — so a slot holding player.json but no timeService.json left self.timeService.player/.stats pointing at the discarded defaults.
• Everything on the daily tick runs off those references: TimeService.increaseDay credits bank interest to self.player, then hands self.player, self.stats to business.runDailyProduction, investments.runDailyIncome and housing.runDailyRent. With a stale reference, a day rollover silently credited interest, crew catch and rental income to an object nobody reads, and charged rent against it too.
• Fixed by rebinding self.timeService.player/.stats once after the load block, alongside the existing self.userInterface.player/.timeService repointing.

That slot shape is reachable in practice: SaveFileManager.migrate_old_save_files gates on player.json only and moves whichever of the three files exist; the same shape also results from a save interrupted partway through FishE.save() (which writes player.json first), or from a zero-byte timeService.json that the getsize(...) > 0 guard skips.

Test plan

• python3 -m compileall -q src tests
• python3 -m pytest --cov=src --cov-report=term-missing --cov-report=xml:cov.xml — 371 passed (368 before), coverage 94%
• All three new tests were confirmed to fail against main (stashing only the src/fishE.py change) and pass with it
• No front-end change — this is game-state wiring in FishE.__init__, below the BaseUserInterface contract, so console/pygame/web are all unaffected
• No Player/Stats/TimeService field added, renamed or retyped, so schemas/*.json and the *JsonReaderWriter classes are unchanged (re-verified all three still match their reader-writer field-for-field)

New tests in tests/test_fishE.py, driving the real FishE.__init__ against a temp save slot with only the save-slot menu and front-end stubbed:

• test_init_rebinds_timeService_to_loaded_player_without_timeService_file — slot with player.json + stats.json, no timeService.json
• test_init_rebinds_timeService_when_only_player_file_present — interrupted-save shape
• test_init_daily_tick_credits_the_loaded_player — behavioural: increaseDay() reaches the player the game actually reads

Closes #120

---

This PR description was drafted during a Gardener session (Stephenson-Software/gardener).

[dmccoystephenson]

Self-review: the fix is correct and minimally scoped. Verified loadPlayer/loadStats/loadTimeService are called only from __init__ (src/fishE.py:57, :61, :65 — no other call sites), so a single rebind after the load block covers every path that can replace self.player/self.stats, including the except fallbacks inside the loaders. It runs unconditionally, so the already-correct case (timeService.json present) re-assigns the same objects and is a no-op. Placement is before the self.locations dict is built, so every location gets the same TimeService instance it always did. Schemas and reader-writers re-checked field-for-field and are untouched by this change; no front-end path is involved. Two notes below, neither blocking.

[dmccoystephenson]

Rebinding attributes here keeps the fix small, but it does leave two places that must agree about what the current player is (this block and the TimeService.__init__ that captured them). A more structural fix would be to defer constructing TimeService until after the load block entirely — it is only needed earlier because UserInterfaceFactory.create_user_interface takes it at line 44 so the save-file menu can render. Not worth the churn in this PR, and the comment above documents the coupling, but worth knowing if the load sequence is reworked later.

[dmccoystephenson]

This helper reassigns fishE's module globals without restoring them, which is the same thing createFishE (line 22) and the existing persistence tests already do in this file, so it is consistent rather than new. It is also order-independent in both directions: createFishE re-mocks every global it cares about, and this helper re-binds every real class it needs, so neither can be poisoned by the other regardless of which runs first. Confirmed by running the full suite (371 passed).