[improve][broker] Improve no-filter dispatch performance by skipping filter calls

[void-ptr974]

Motivation

AbstractBaseDispatcher#filterEntriesForConsumer is on the dispatch hot path. When no entry filters are configured, runFiltersForEntry only returns ACCEPT, so calling it once per entry is avoidable work.

Modifications

• Cache the dispatcher filter state in a local variable for the batch.
• Skip runFiltersForEntry when no entry filters are configured.

Scope

This intentionally leaves the filtered/redelivery list allocation and ack-set handling unchanged. Those are separate optimizations and are kept out of this PR to make the review focused on the no-filter fast path only.

Performance

This improves no-filter dispatch performance by removing one no-op filter call and the REJECT/RESCHEDULE checks per entry when no filters are configured. Allocation remains effectively unchanged; this is a CPU-path optimization.

JMH data gathered locally using a local-only DispatcherDispatchPathBenchmark.(noFilterRunFiltersPerEntry|noFilterFastPath) benchmark, with -p entriesCount=32,256,1000 -p batchSize=1,10 -wi 2 -i 5 -w 1s -r 1s -f 1 -bm avgt -tu ns -prof gc:

batchSize	entriesCount	old path	fast path	improvement
1	32	44.8 ns/op	28.0 ns/op	37.6%
1	256	329.1 ns/op	186.1 ns/op	43.4%
1	1000	3151.4 ns/op	1603.3 ns/op	49.1%
10	32	52.4 ns/op	31.2 ns/op	40.4%
10	256	361.7 ns/op	182.9 ns/op	49.4%
10	1000	2658.4 ns/op	1593.4 ns/op	40.1%

Verification

• Added AbstractBaseDispatcherTest coverage for a 512-entry no-filter dispatch batch with varying payload sizes and batch sizes.
• ./gradlew :pulsar-broker:compileJava :pulsar-broker:checkstyleMain --max-workers=1
• ./gradlew :pulsar-broker:test --tests org.apache.pulsar.broker.service.AbstractBaseDispatcherTest --max-workers=1 -PtestRetryCount=0
• ./gradlew :pulsar-broker:checkstyleTest --max-workers=1