PHOENIX-7815 Make WAL sync durability configurable

[tkhurana]

Summary

• Make the replication log writer's sync durability configurable via the existing hbase.wal.hsync HBase property (default false, matching HRegion.DEFAULT_WAL_HSYNC).
• When hbase.wal.hsync=true, sync() calls hsync() (forces fsync to disk). When false (the default), it calls hflush() (flushes to datanode memory only), trading durability for lower latency.
• Plumbs the flag through LogFileWriterContext into both HDFSDataOutput and AsyncFSDataOutput, which previously hard-coded hsync() in sync().

Changes

• LogFileWriterContext: read hbase.wal.hsync from config (defaults to HRegion.DEFAULT_WAL_HSYNC = false); add getUseHsync()/setUseHsync(); include useHsync in toString().
• HDFSDataOutput / AsyncFSDataOutput: accept useHsync in constructor; sync() dispatches to hsync() or hflush() based on the flag.
• LogFileWriter: propagate context.getUseHsync() when constructing HDFSDataOutput.
• Tests: LogFileFormatTest updated for the new constructor arg; LogFileWriterSyncTest adds testSyncWithHflush and sets hbase.wal.hsync=true explicitly in the existing hsync test.

Test plan

• mvn test -pl phoenix-core -Dtest=LogFileWriterSyncTest — 5/5 pass
• mvn test -pl phoenix-core -Dtest=LogFileFormatTest — 16/16 pass
• mvn spotless:apply clean

[lokiore]

LGTM