fix(miner): retain unmarked send cache

[JSONbored]

Summary

• retain unmarked destination-send cache entries when cleanup sees them as stale
• keep marked stale entries removable so completed local cache still clears normally
• add regression coverage for poller misses, cleanup, rediscovery, and mark_fulfilled retry behavior

Why

A transient contract/indexer read gap can make the miner poller drop an active swap after repeated get_swap() misses. The miner then calls cleanup_stale_sends() with the reduced active set. If that cleanup deletes an unmarked cached destination tx, a later rediscovered swap can broadcast destination funds again instead of retrying mark_fulfilled for the original tx.

Closes #353.

Validation

• uv run pytest tests/test_fulfillment.py tests/test_swap_poller.py -q
• uv run pytest tests/ -q
• uv run ruff check .
• uv run ruff format --check .
• git diff --check
• Codex Security diff scan: no reportable findings
• CodeRabbit CLI review: 0 issues

Notes

The duplicate check found no open issue or PR for this exact cleanup_stale_sends / unmarked send-cache failure mode before publication.

[LandynDev]

Can you give this a quick rebase on entrius:test and I'll dive in on reviewing it/checking it out @JSONbored

[JSONbored]

Can you give this a quick rebase on entrius:test and I'll dive in on reviewing it/checking it out @JSONbored

Done! Thanks 🙌🏻