ci: run the scheduled dependency check daily

[lesnik512]

scheduled.yml moves from 0 6 * * 1 to 0 6 * * *.

Why

CI has no committed lockfile (uv.lock is gitignored), so every run resolves dependencies fresh
against the live index. That makes the scheduled job's only distinct value its timer: it finds a
breaking release before a contributor trips over it. A weekly cadence puts up to seven days between
the release and the report.

That window was not hypothetical here. faststream 0.7.6 shipped, broke ty check and five tests,
and was found by hand in #182 rather than by the cron. Tomorrow's Monday run would have caught it,
a day late and by coincidence.

The repo is public, so the extra runs consume no Actions minutes.

Noise

The failure reporter already dedupes: the first failure opens one issue, subsequent failures comment
on it. Daily means a persistent failure adds a comment a day instead of one a week, which is the
intended signal rather than noise. A fixed failure stops immediately.

Also dropped "weekly" from the scheduled-failure label description and the issue body, which the
schedule no longer matched. gh label create --force in the reporter updates the live label
description on the next failing run.

Not changed

The check itself. _checks.yml is shared with PR CI and is untouched.

[github-actions]

Benchmark gate

✅ gate passed

scenario	msg/s	delete/msg	WALrec/msg	WALB/msg	fpi	upd	del	dead_tup
consumer/w1/b10	957	1.000	6.80	918	242	5000	5000	10000
consumer/w1/b100	964	1.000	6.84	918	243	5000	5000	10000
consumer/w2/b10	1132	1.000	6.80	962	242	5000	5000	10000
consumer/w2/b100	1322	1.000	6.74	958	243	5000	5000	10000
consumer/w4/b10	1029	1.000	6.82	995	262	5000	5000	10000
consumer/w4/b100	1534	1.000	6.86	1036	244	5000	5000	10000
consumer/w1/b100/tfbs100	5333	0.010	6.07	1116	243	5000	5000	10000
producer/w1/b100	2095	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).