fix(inkless:retention): size retention enforcement to keep up with produce

[jeqo]

The defaults could not keep up with a continuously written partition. Sustained deletion capacity per partition is max.batches.per.request per retention.enforcement.interval.ms, i.e. 1000 batches per 5 min = 3.3 batches/s. A partition gets roughly one batch per produce.commit.interval.ms (250ms) per broker writing to it, so 4 batches/s from a single writer already exceeded the delete rate, and in diskless any broker can accept writes for any partition. The batches table therefore grew without bound on busy partitions while data stayed past its retention.

Raise capacity by ~10x:

• retention.enforcement.interval.ms: 5 min -> 1 min
• retention.enforcement.max.batches.per.request: 1000 -> 2000

That is 33 batches/s per partition, which covers a partition written by ~8 brokers at the default commit interval.

The interval was the better of the two levers to move. Capacity is cap/interval, so it could also have come from a 10x cap, but the per-pass boundary scan is O(cap) (enforce_retention_v2 after V22) and the delete holds the log's row: fewer, heavier passes hold locks longer, which is what V21/V22 set out to reduce. More frequent, bounded passes keep the lock hold short. The cost is transaction rate: every due partition costs one transaction whether or not anything expires, so enforcement transactions grow 5x (~330/s -> ~1670/s cluster-wide at 100k partitions).

consolidation.cleanup.interval.ms no longer derives its default from the retention interval, so this change does not silently make WAL pruning 5x more frequent too. Its cost profile is different (whole-file scans, not per-partition boundary scans), so it keeps 5 min explicitly.

Not addressed here: enforcement cadence per broker is interval * brokerCount while producer fan-in per partition grows with brokerCount, so the shortfall is quadratic in cluster size and no fixed default covers every size. Scaling the batch budget with brokerCount (as the scheduler already does for cadence) is the follow-up.