fix: ruff 0.17 ICN001 and batch_bodies setter (ty 0.0.86)

[lesnik512]

Fixes #207.

The scheduled dependency check picked up ruff 0.17.0 and ty 0.0.86, and both fail on main.

• ruff 0.17 now enforces ICN001 for datetime, so import datetime as _dt is renamed to dt in 14 files. This is the ruff autofix. Nothing else in the codebase is named dt, so the rename can't shadow anything.
• ty 0.0.86 reports that OutboxPublishCommand.batch_bodies overrides upstream's writable BatchPublishCommand.batch_bodies with a read-only property. That is a real break: cmd.batch_bodies = ... raised AttributeError. The new setter matches upstream: the first item goes to body and the rest to extra_bodies. An empty sequence sets body=None, so the getter returns (None,), consistent with this class keeping None bodies.

The new test test_publish_command_batch_bodies_is_assignable failed before the fix. Locally, just lint-ci is clean and the full suite passes against Postgres 17 (626 passed).

[github-actions]

Benchmark gate

✅ gate passed

scenario	msg/s	delete/msg	WALrec/msg	WALB/msg	fpi	upd	del	dead_tup
consumer/w1/b10	932	1.000	6.80	918	242	5000	5000	10000
consumer/w1/b100	582	1.000	6.84	920	243	5000	5000	10000
consumer/w2/b10	946	1.000	6.84	1000	261	5000	5000	10000
consumer/w2/b100	902	1.000	6.73	958	244	5000	5000	10000
consumer/w4/b10	832	1.000	6.74	947	242	5000	5000	10000
consumer/w4/b100	1339	1.000	6.86	1032	243	5000	5000	10000
consumer/w1/b100/tfbs100	2906	0.010	6.17	1170	269	5000	5000	10000
producer/w1/b100	3095	0.000	3.04	584	0	0	0	0

Gated (fails the build): delete_calls + tuple counters (upd/del/ins) + the producer's insert_calls, exact; select_calls within +2; wal_records within a 10% band. msg/s, WAL bytes and total calls are informational (timing/FPI noise).